Download EXHAUSTIBLE SETS IN HIGHER

EXHAUSTIBLE SETS IN HIGHER-TYPE COMPUTATION
MARTÍN ESCARDÓ
School of Computer Science, University of Birmingham, B15 2TT, UK
Abstract. We say that a set is exhaustible if it admits algorithmic universal quantification
for continuous predicates in finite time, and searchable if there is an algorithm that, given
any continuous predicate, either selects an element for which the predicate holds or else
tells there is no example. The Cantor space of infinite sequences of binary digits is known
to be searchable. Searchable sets are exhaustible, and we show that the converse also holds
for sets of hereditarily total elements in the hierarchy of continuous functionals; moreover,
a selection functional can be constructed uniformly from a quantification functional. We
prove that searchable sets are closed under intersections with decidable sets, and under
the formation of computable images and of finite and countably infinite products. This
is related to the fact, established here, that exhaustible sets are topologically compact.
We obtain a complete description of exhaustible total sets by developing a computational
version of a topological Arzela–Ascoli type characterization of compact subsets of function
spaces. We also show that, in the non-empty case, they are precisely the computable
images of the Cantor space. The emphasis of this paper is on the theory of exhaustible
and searchable sets, but we also briefly sketch applications.
1. Introduction
A wealth of computational problems of interest have the following form:
Given a set K and a property p of elements of K, decide whether or not all
elements of K satisfy p.
For K fixed in advance, this is equivalent to the emptiness problem for p. One is often
interested in suitable restrictions on the possible syntactical forms of the predicate p that
guarantee that this problem is decidable (or, less ambitiously, that the non-emptiness problem is semi-decidable) uniformly in the syntactical form of p. In this work, on the other
hand, the emphasis is on the set K rather than the predicate p, and we study the case in
which K is infinite. Moreover, p is not assumed to be given syntactically or via any other
kind of intensional information: we only use information about the input-output relation
determined by p considered as a boolean-valued function. In the absence of intensional
information, continuity of p plays a fundamental role, where p is continuous iff for any x
in the domain of p, the boolean value p(x) depends only on a finite amount of information
2000 ACM Subject Classification: F.4.1, F.3.2.
Key words and phrases: Higher-type recursion, continuous functional, PCF, domain theory, Scott domain,
semantics, topology, compactly generated space, functional programming, Haskell.
2000 Mathematics Subject Classification: 03D65, 68Q55, 06B35, 54D50.
LOGICAL METHODS
IN COMPUTER SCIENCE
DOI:10.2168/LMCS-???
1
c Escardó
Creative Commons
2
ESCARDÓ
about x. We work in the realm of higher-type computation with continuous functionals, using Ershov–Scott domains to model partial functionals, and Kleene–Kreisel spaces to model
total functionals [34, 31].
We say that the set K is exhaustible if the above problem can be algorithmically solved
for any continuous p defined on K, uniformly in p. The uniform dependency on p is
formulated by giving the algorithm the type (D → B) → B, where D is a domain, K ⊆ D,
and B is the domain of booleans. The main question investigated in this work is what kinds
of infinite sets are exhaustible.
Clearly, finite sets of computable elements are exhaustible. What may be rather unclear
is whether there are infinite examples. Intuitively, there can be none: how could one
possibly check infinitely many cases in finite time? This intuition is correct when K is a
set of natural numbers: it is a theorem that, in this case, K is exhaustible if and only if it
is finite. This can be proved by reduction to the halting problem, but there is also a purely
topological argument (Remark 5.6). However, it turns out that there is a rich supply of
infinite exhaustible sets. A first example, the Cantor space of infinite sequences of binary
digits, goes back to the 1950’s, or even earlier, with the work of Brouwer, as discussed in
the related-work paragraph below.
We say that K is searchable if there is an algorithm that, given any continuous predicate p, either selects some x ∈ K such that p(x) holds, or else reports that there isn’t
any. It is easy to see that searchable sets are exhaustible. We show that, for sets of total
elements, the converse also holds and hence the two notions coincide. Moreover, a selection
functional can be constructed uniformly from a quantification functional (Section 6).
We develop tools for systematically building exhaustible and searchable sets, and some
characterizations, including the following: they are closed under intersections with decidable
sets, under the formation of computable images and of finite and countably infinite products
(Section 4). In the case of exhaustibility, the last claim is restricted to sets of total elements,
and is open beyond this case. The non-empty exhaustible sets of total elements are precisely
the computable images of the Cantor space (Section 6). We also formulate and prove an
Arzela–Ascoli type characterization of exhaustible sets of total elements of function types
(Section 7).
The above closure properties and characterizations resemble those of compactness in
topology. This is no accident: we show that exhaustible sets of total elements are indeed
compact, in the Kleene–Kreisel topology (Section 5). This plays a crucial role in the correctness proofs of some of the algorithms, and, indeed, in their very construction. Thus,
the specifications of all of our algorithms can be understood without much background,
but an understanding of the working of some of them requires some amount of topology.
We have organized the presentation so that the algorithms occurring earlier are motivated
by topology but don’t rely on knowledge of topology for their formulation or correctness
proofs.
In Section 2 we include background material that can be consulted on demand and in
Section 3 we define the central notions investigated in this work. In Section 8 we include
technical remarks, further work, announcement of results, applications, and research directions. In the concluding Section 9 we review the role topology plays in our investigation of
exhaustible and searchable sets.
EXHAUSTIBLE SETS
3
Related work. Brouwer’s Fan functional gives the modulus of uniform continuity of a discretevalued continuous functional on the Cantor space. According to personal communication
by Dag Normann, computability of the Fan functional was known in the late 1950’s. This
immediately gives rise to the exhaustibility of the Cantor space. A number of authors have
considered the definability of the Fan functional in various formal systems. Normann [34]
cites Tait (1958, unpublished), Gandy (around 1982, unpublished) and Berger [7] (1990).
Tait showed that the Fan functional is not definable from Kleene’s schemes S1–S9 interpreted over total functionals. Berger observed that, for partial functionals, PCF definability
coincides with S1–S9 definability, and showed that the Fan functional is PCF definable. In
order to do that, he first explicitly defined a selection functional for the Cantor space.
Then Hyland informed the community that Gandy was aware of the PCF/S1–S9 definability of the Fan functional for the partial interpretation of Kleene’s schemes, but Gandy’s
construction seems to be lost.
Acknowledgements. I have benefited from stimulating discussions with, and questions by,
Andrej Bauer, Ulrich Berger, Dan Ghica, Achim Jung, John Longley, Paulo Oliva, Matthias
Schröder, and Alex Simpson. I also thank Dag Normann for having answered many questions regarding the history and technical ramifications of the subject of higher-type computation, and for sending me a copy of Tait’s unpublished manuscript — but the reader
should consult his paper [34] for a more accurate and detailed account.
2. Background
The material developed here can consulted on demand, except for Section 2.1, which
introduces and briefly discusses our model of computation. Some readers will be more familiar with domain theory and Ershov–Scott continuous functionals (and PCF or functional
programming) via denotational semantics, and others with the Kleene–Kreisel continuous
functionals via higher-type computability theory, and we consider these two models and their
relationship [34]. Alternatively, we could have worked with Weihrauch’s model of computation via representations [46], which generalizes Kleene’s approach via associates [31]. Even
better, we could have worked with the QCB model of computation, which subsumes both
domain theory and representation theory in a natural way [3, 4]. We adopt the Ershov–Scott
and Kleene–Kreisel approaches as they have played a wide role [34]. A presentation based
on QCB spaces would have been not only more general but also cleaner in several ways,
but less familiar and perhaps more technically demanding. Our objective in this paper is to
address the essential issues without getting distracted by an excessive amount of generality.
2.1. Domains of computation. We work on a cartesian closed category of computable
maps of effectively given domains that contains the flat domains of booleans,
B = {0, 1, ⊥},
and of natural numbers,
N = N ∪ {⊥},
such as [17] or [44] among other possibilities. Notice that these categories are closed under
countable cartesian powers. We don’t need to, and we don’t, explicitly refer to effective
presentations, and in particular to numberings of finite elements or abstract bases etc. to formulate computability results. We instead start from well known computable functions and
4
ESCARDÓ
use the fact that computable functions are closed under definition by lambda abstraction,
application, least fixed points etc. Moreover, we don’t invoke non-sequential functions such
as Platek’s parallel-or [41] or Plotkin’s parallel-exists [35] in order to construct new computable functions. Our algorithms can thus be directly understood as functional programs
in e.g. PCF [35], FPC [37] (PCF extended with recursive types, interpreted as solutions of
domain equations) or practical versions of FPC such as Haskell [12, 25], as done in [19].
At some point we need further assumptions on our domains of computation to be
able to formulate and prove certain results. Some of those results can be formulated, and
perhaps also be proved, for domains with totality in the sense of Berger [8]. We consider
the particular case consisting of the smallest collection of domains containing B and N and
closed under finite products, countable powers and exponentials (=function spaces).
2.2. Higher-type computation. The remainder of this section is not needed until Theorem 4.9. As discussed in e.g. [34, 29, 30], there are many approaches to higher-type
computation. Kleene defined the total functionals directly, but it has been found more convenient to work with the larger collection of partial functionals and isolate the total ones
within them, as done by Kreisel. The approaches are equivalent, and such total functionals
are often referred to as Kleene–Kreisel functionals or continuous functionals. It turns out
that, as discussed by Normann [34], this coincides with another approach that also arises
programming language semantics: equivalence classes of total functionals on Ershov–Scott
domains. We work with both total functionals on domains and a characterization of the
Kleene–Kreisel functionals, due to Hyland, in terms of compactly generated spaces.
Types. The simple types are defined by induction as
σ, τ ::= o | ι | σ × τ | σ → τ,
where o and ι are ground types for booleans and natural numbers respectively. The subset
of pure types is defined by
σ ::= ι | σ → ι.
As usual, we’ll occasionally reduce statements about simple types to statements about pure
types.
Partial functionals. For each type σ, define a domain Dσ of partial functionals of type σ by
induction as follows:
Do = B, Dι = N ,
Dσ×τ = Dσ × Dτ ,
Dσ→τ = (Dσ → Dτ ) = Dτ Dσ
where the products and exponentials are calculated in the cartesian closed category of
continuous maps of Scott domains, where a Scott domain is an algebraic, bounded complete,
and directed complete poset [1].
EXHAUSTIBLE SETS
5
Total functionals. For each type σ, define a set Tσ ⊆ Dσ of total functionals and a relation ∼σ on Dσ as follows, where γ ranges over the ground types o and ι:
To = 2 = {0, 1},
x ∼γ y ⇐⇒ x, y ∈ Tγ ∧ x = y.
Tι = N,
Tσ×τ = Tσ × Tτ ,
(x, x0 ) ∼σ×τ (y, y 0 ) ⇐⇒ x ∼σ y ∧ x0 ∼τ y 0 ,
Tσ→τ = {f ∈ Dσ→τ | f (Tσ ) ⊆ Tτ },
f ∼σ→τ g ⇐⇒ ∀x ∼σ y.f (x) ∼τ g(y).
Then the set Tσ can be recovered from the relation ∼σ as
x ∈ Tσ ⇐⇒ x ∼σ x,
and the relation can be recovered from the set as
x ∼σ y
⇐⇒
⇐⇒
x u y ∈ Tσ
x, y ∈ Tσ and x and y are bounded above.
See e.g. [8] and [38]. In particular, ∼σ is an equivalence relation on Tσ .
Computability. Plotkin [35] characterized the computable partial functionals as those that
are PCF-definable from parallel-or and parallel-exists [35]. All computable functionals we
construct from Section 6 onwards are defined in PCF without parallel extensions. This
characterization of computability includes, in particular, total functionals. An interesting
fact, which we don’t need to invoke, is that every total functional definable in PCF with
parallel extensions is equivalent to one definable in PCF without parallel extensions [32].
2.3. Kleene–Kreisel functionals. For each type σ, define by induction a set Cσ of
Kleene–Kreisel functionals of type σ and a surjection ρσ : Tσ → Cσ as follows, so that
Cσ ∼
= Tσ / ∼σ .
For ground types and product types, define
Co = 2 = To , Cι = N = Tι ,
Cσ×τ = Cσ × Cτ ,
ργ (x) = x.
ρσ×τ = ρσ × ρτ .
For function types, consider the diagram
Dσ (†)
f
⊃
Tσ
(1)
- Cσ
(2)
?
Dτ ρσ
?
⊃
Tτ
ρτ -
φ
?
Cτ .
The square (1) commutes for some map Tσ → Tτ if and only if f ∈ Tσ→τ , and in this case
the map is uniquely determined as the (co)restriction of f . Moreover, in this case, there is
a unique map φ making the square (2) commute, because ρσ is a surjection. We define
Cσ→τ = {φ : Cσ → Cτ | ∃f ∈ Tσ→τ .(2) commutes},
ρσ→τ (f ) = the unique φ such that (†) commutes.
Then, by construction, for any σ and all x, y ∈ Dσ , we have that
x ∼σ y iff x, y ∈ Tσ and ρσ (x) = ρσ (y).
6
ESCARDÓ
If (†) commutes, we say that f is a representative of φ. A Kleene–Kreisel functional is
computable iff it has a computable representative.
Lemma 2.1. Every Cσ is a computable retract of Cτ →ι for some τ .
A stronger form of this is known as “simple types are retracts of pure types” (see
e.g. [30]). Here we use the fact that every pure type is either ι or of the form τ → ι, and
that ι is a retract of τ → ι for any τ .
2.4. Compactly generated spaces. The remainder of this section is not needed until
Section 5, where it is used in order to formulate and prove the crucial Lemma 5.5 that establishes compactness of exhaustible sets of total elements. Compactly generated Hausdorff
spaces, or k-spaces, to be introduced shortly, are to total computation as Scott domains
are to partial computation. This becomes clear in Section 2.5. Here we briefly introduce
k-spaces and some of their fundamental topological properties that are applied to prove
various computational theorems. For more details and proofs, see e.g. [21] or the references
contained therein.
We begin by considering the Hausdorff case. If F is a closed set of a Hausdorff space X,
then K ∩ F is closed for every compact set K ⊆ X. Any set F that satisfies this condition
is called k-closed. The Hausdorff space X is a k-space iff every k-closed set is closed. (This
is equivalent to saying that X is the colimit of its compact subspaces ordered by inclusion.)
Any Hausdorff space can be transformed into a k-space by stipulating that all k-closed
sets are closed. In categorical terms, this construction is a coreflection of the category of
Hausdorff spaces into its subcategory of k-spaces.
The category of Hausdorff k-spaces is cartesian closed. Given objects X and Y , their
categorical product X ×Y is the coreflection of their topological product. Their exponential
Y X consists of the continuous maps X → Y under the coreflection of the compact-open
topology. The compact-open topology has subbasic open sets of the form
N (K, V ) = {f ∈ Y X | f (K) ⊆ V },
where K is a compact subset of X and V is an open subset of V . Cartesian closedness
amounts to the fact that the evaluation map
YX ×X → Y
(f, x) 7→ f (x)
is continuous, and that for any continuous map f : Z × X → Y , its transpose
f¯: Z → Y X
z 7→ (x 7→ f (z, x))
is continuous. Equivalently, f is continuous iff f¯ is continuous.
We also need to consider k-spaces without the restriction to the Hausdorff case. Let
X be an arbitrary topological space. A probe is a continuous function p : K → X where
K is a compact Hausdorff space. A set F ⊆ X is k-closed if p−1 (F ) is closed for every
probe p : K → X. Then, again, X is a k-space iff every k-closed set is closed, and kspaces form a cartesian closed category, and the inclusion of Hausdorff k-spaces preserves
products and exponentials. All locally compact spaces are k-spaces, and this includes nonHausdorff examples such as Scott domains under the Scott topology. The description of the
products and exponentials in the general case is omitted and the reader is referred to the
EXHAUSTIBLE SETS
7
above references, but in any case they are not needed for the purposes of this work, with
one exception: an exponential of k-spaces whose base is the Sierpinski space has the Scott
topology [21]. This is applied in the proof of the following lemma.
Denote by S the Sierpinski space with an isolated point > and a limit point ⊥. This
is the same as the domain {⊥, >} under the Scott topology. For any topological space X,
a function p : X → S is continuous iff p−1 (>) is open, and a set U ⊆ X is open iff its
characteristic function χU , defined by χU (x) = > ⇐⇒ x ∈ U , is continuous. Thus, using
the Sierpinski space, the notion of openness is reduced to that of continuity. The following
reduces the notion of compactness to that of continuity (a particular case of this is proved
in [18], with essentially the same proof as the one give here).
Lemma 2.2. If X is a k-space, a set K ⊆ X is compact if and only if the universal
quantification functional ∀K : S X → S defined by
∀K (p) = > iff p(x) = > for all x ∈ K
is continuous.
Proof. A set K is compact if and only if every directed cover of K by open sets has a
member that covers K, because from any cover one obtains a directed cover with the same
union by adding the finite unions of the members of the cover. Hence by definition of the
Scott topology, a set is compact if and only if its open-neighbourhood filter is open in the
Scott topology of the lattice of open sets. But U 7→ χU is a bijection from the lattice of open
sets to the points of S X , and it was shown in [21] that the topology of the exponential S X is
the one induced by this bijection. Hence the functional ∀K is continuous iff ∀−1
K (>) is open
iff the set of characteristic functions χU with K ⊆ U is open iff the open neighbourhood
filter of K is open iff K is compact.
2.5. Hyland’s characterization of the Kleene–Kreisel functionals. For certain constructions and proofs of algorithms, we consider a topology on the set of Kleene–Kreisel
functionals.
Definition 2.3. Endow Tσ with the relative Scott topology and Cσ with the quotient
topology of the surjection ρσ : Tσ → Cσ . We refer to this topology on Cσ as the Kleene–
Kreisel topology, and to the resulting spaces Cσ as the Kleene–Kreisel spaces. The points of
the Kleene–Kreisel spaces are often referred to as the continuous functionals in the highertype computability (or higher-type recursion) literature.
A proof of the following inductive topological characterization of the Kleene–Kreisel
spaces, attributed to Hyland, can be found in Normann [31].
Lemma 2.4. With products and exponentials in the category of Hausdorff k-spaces,
(1) Cγ has the discrete topology for γ ground,
(2) Cσ×τ = Cσ × Cτ and
(3) Cσ→τ = Cτ Cσ .
The following two lemmas, which are part of the folklore of the subject, are applied
in order to show that exhaustible sets of total elements are compact in the Kleene–Kreisel
topology (Lemma 5.5(1)). A set is called clopen if it is both closed and open.
Lemma 2.5. For every clopen U ⊆ Cσ there is a total predicate p ∈ (Dσ → B) such that
−1
−1
−1
ρ−1
σ (U ) ⊆ p (1) and ρσ (Cσ \ U ) ⊆ p (0).
8
ESCARDÓ
Proof. Because U is clopen, its characteristic function χU : Cσ → 2 is continuous, and hence
so is the composite i ◦ χU ◦ ρσ : Tσ → B, where i : 2 → B in the inclusion. Because T is dense
in Dσ (see e.g. [8]) and because Scott domains, and hence B, are densely injective (see e.g.
[24]), by definition of injectivity this extends to a continuous function p : Dσ → B. Then p
is total by construction, and the extension property amounts to the above set inclusions.
A space is zero-dimensional iff it has a base of clopen sets. The zero-dimensional
reflection ZC of a space C is obtained by taking the same set of points and the clopen sets
as a base.
Lemma 2.6. ZCσ and Cσ have the same compact subsets.
Proof. We first show that KZC = C where C = Cσ and K is the coreflector into the
category of k-spaces. The property KZC = C is easily seen to be inherited by retracts,
and hence, by Lemma 2.1, it is enough to consider σ = τ → ι, and hence C = NY for some
k-space Y . Exponentials in k-spaces are given by the k-coreflection of the compact-open
topology on the set of continuous maps. When the target is N, the compact-open topology
is clearly zero-dimensional and Hausdorff. Now, it is easy to see that KZC = C iff there is
some zero-dimensional topology whose k-reflection is C, and hence we are done. The result
then follows from the well-known fact that a Hausdorff space has the same compact sets as
its k-coreflection.
If the spaces Cσ were zero-dimensional, the above lemma would be superfluous. But
Matthias Schröder [39] has recently shown, after this paper was produced and refereed, that
the spaces Cσ are not zero-dimensional, and in fact not even regular, answering a question
of [5, 33].
3. Exhaustible and searchable sets
We now formulate the central notions investigated in this work.
Definition 3.1. if K is a subset of the domain D, we say that a predicate p ∈ (D → B) is
defined on K if p(x) 6= ⊥ for every x ∈ K.
Definition 3.2. We say that a subset K of the domain D is exhaustible if there is a
computable functional ∀K : (D → B) → B such that for any p ∈ (D → B) defined on K,
(
1 if p(x) = 1 for all x ∈ K,
∀K (p) =
0 if p(x) = 0 for some x ∈ K.
Such a universal quantification functional is not uniquely determined, because its behaviour
is not specified for predicates p that are not defined on K. For the sake of clarity, we’ll
often write “∀K (λx. . . . )” as “∀x ∈ K. . . . ”.
Clearly, it is equivalent to instead require the existence of a computable functional
∃K : (D → B) → B such that for any p ∈ (D → B) defined on K,
(
1 if p(x) = 1 for some x ∈ K,
∃K (p) =
0 if p(x) = 0 for all x ∈ K,
because such functionals are inter-definable by the De Morgan Laws and hence we’ll freely
switch between them.
EXHAUSTIBLE SETS
9
We now formulate searchability in a way slightly different from that of the introduction,
which is more convenient for our purposes. The only essential difference is that the present,
official definition excludes the empty set (cf. Remark 3.5).
Definition 3.3. We say that a set K ⊆ D is searchable if there is a computable functional
εK : (D → B) → D such that, for every predicate p ∈ (D → B) defined on K,
(1) εK (p) ∈ K, and
(2) p(εK (p)) = 1 if p(x) = 1 for some x ∈ K.
Again, notice that the selection functional εK is not uniquely determined by K.
Thus, εK (p) is an example of an element of K for which p holds, if such an element
exists, or a counter-example in K if no such example exists.
Lemma 3.4. Searchable sets are exhaustible.
Proof. Define ∃K (p) = p(εK (p)).
The empty set is exhaustible with ∀∅ (p) = 1, but it is not searchable because the
condition ε∅ (p) ∈ ∅ cannot hold. But we’ll see in Section 6 that, under fairly general and
natural conditions, the two notions turn out to agree in the non-empty case. Moreover, it is
clear that non-empty finite sets of computable elements are both exhaustible and searchable.
Remark 3.5. With 1 = {?}, an equivalent definition of searchability is that
(1) K has a computable element eK , and
(2) there is ε0K : (D → B) → 1 + D computable such that ε0K (p) = ? if there is no
example, and otherwise ε0K (p) ∈ K and p(ε0K (p)) = 1.
In fact, given εK one can define eK = εK (λx.1) and
ε0K (p) = if p(εK (p)) then εK (p) else ?.
Conversely, given ε0K and eK as specified, one can define
εK (p) = if ε0K (p) = ? then eK else ε0K (p).
Regarding examples, we’ll deduce later the known fact that the Cantor space is searchable. For the moment, we show that the natural numbers with a point at infinity form a
searchable set.
Definition 3.6. The one-point compactification of the natural numbers is the subspace
N∞ of the Cantor space 2ω ⊆ B ω consisting of the sequences 0n 1ω (representing natural
numbers n) and 0ω (representing the added point at infinity).
The relative Scott topology on the Cantor space agrees with the product topology of
the discrete space 2, but such topological considerations are not needed until Section 5. In
constructive mathematics, N∞ is equivalently defined as the set of sequences α ∈ 2ω with
αi ≤ αi+1 , to avoid excluded middle. In functional programming, N∞ also arises as the set
of maximal elements of the domain of lazy natural numbers.
Example 3.7. N∞ is searchable, with selection functional εN∞ defined by primitive recursion as
εN∞ (p)(i) = ∃n ≤ i. p(0n 1ω ).
Notice that εN∞ (p) is the infimum of the set of solutions α ∈ N∞ of p(α) = 1, including the
case in which the set is empty, for which εN∞ (p) = ∞.
10
ESCARDÓ
This construction is implicit in Exercise 1 of Barendregt [2, Page 581], attributed to
Kreisel. The point of that exercise is that this algorithm can be interpreted as a functional
in the full type hierarchy, defined in Gödel’s system T , that also works for discontinuous p.
The exercise uses this to prove that the substructure of definable elements is not extensional,
or equivalently, that the set-theoretical model of system T fails to be fully abstract. This
exercise was brought to my attention by Gordon Plotkin and Alex Simpson, after I posed
this full abstraction question to them. Notice that the Kleene–Kreisel model of system T
is fully abstract, using the fact that the elements of a dense set are definable.
4. Building new searchable sets from old
In this section we develop algorithms that don’t require knowledge of topology but
are motivated by topological considerations. Starting from the finite sets, the algorithms
allow us to systematically build plenty of infinite searchable sets. The intuition behind the
topological notion of compactness is that compact sets behave, in many relevant respects,
as if they were finite. Infinite sets that admit exhaustive search in finite time share the same
intuition. Hence it is natural to conjecture that they also share similar structural properties.
For example, compact sets are closed under the formation of products (Tychonoff theorem).
Motivated by this, in this section we show that searchable sets are closed under countable
products, and we also export other closure properties from topology to computation.
Definition 4.1. For a given set K ⊆ D, we say that a set F ⊆ K is decidable on K if
there is a computable map ψF : D → B defined on K such that, for all x ∈ K, ψF (x) = 1
iff x ∈ F .
Proposition 4.2. Let K ⊆ D and let F ⊆ K be decidable on K.
(1) If K is exhaustible then so is K ∩ F .
(2) If K is searchable then so is K ∩ F , provided it is non-empty.
Proof. Define
∃K∩F (p) = ∃x ∈ K.x ∈ F ∧ p(x),
εK∩F (p) = if ∃x ∈ K. x ∈ F ∧ p(x) then εK (λx.x ∈ F ∧ p(x)) else εK (λx.x ∈ F ).
The topological motivation for the above proposition is that the intersection of a closed
set with a compact set is compact. Decidable sets correspond to sets that are open and
closed, and hence, bearing in mind that exhaustible sets (ought to) correspond to compact
sets, the above proposition ought to be true, which it is. It is an easy exercise to show that
exhaustible and searchable sets are closed under binary unions. But binary intersections
are problematic. In fact, in topology, in the absence of assumptions such as the Hausdorff
separation axiom, compact sets fail to be closed under binary intersections. Hence any
algorithm for binary intersections would have to exploit specialized topological and/or ordertheoretic properties of domains. The topological motivation for the following proposition
is that, in topology, continuous images of compact sets are compact. In fact, it arises by
replacing continuity by computability and compactness by exhaustibility.
Proposition 4.3. Exhaustible and searchable sets are closed under the formation of computable images.
EXHAUSTIBLE SETS
11
Proof. Let f ∈ (D → D0 ) be computable and let K be a subset of D. For any quantification
functional ∀K : (D → B) → B, the functional ∀f (K) : (D0 → B) → B defined by
∀f (K) (q) = ∀x ∈ K.q(f (x))
is clearly a quantification functional for f (K).
For any selection functional εK : (D → B) → D, the following definition gives a selection
functional εf (K) : (D → B) → D:
εf (K) (q) = f (εK (λx.q(f (x))).
That is, first find x such that q(f (x)) holds, using εK , and then apply f to this x.
The following corresponds to the fact that compact sets in topology are closed under
finite products:
Proposition 4.4. Exhaustible and searchable sets are closed under the formation of finite
products.
Proof. For K ⊆ D and K 0 ⊆ D0 exhaustible, define
∀K×K 0 (p) = ∀x ∈ K.∀x0 ∈ K 0 .p(x, x0 ).
For K ⊆ D and K 0 ⊆ D0 searchable, to compute εK×K 0 (p) we first find x ∈ K such that
there is x0 ∈ K 0 with p(x, x0 ), and then find x0 ∈ K 0 such that p(x, x0 ), i.e.
x = εK (λx.∃x0 ∈ K 0 .p(x, x0 )),
x0 = εK 0 (λx0 .p(x, x0 )),
using the fact that searchable sets are exhaustible, and let εK×K 0 (p) = (x, x0 ).
Compact sets in topology are closed under arbitrary products. We now show that
searchable sets are closed under countable products. We
like to show that for any
Q would Q
sequence of searchable sets Ki ⊆ Di , their product i Ki ⊆ i Di is also searchable,
but this would require dependent types, which are not part of the traditional higher-type
computation formalism (but see [10] and [9]). So we assume
Q that the components Ki of the
product are all subsets of the same domain D, so that i Ki ⊆ Dω instead, leaving the
more general question for future work.
Given selection functionals
εKi ∈ ((D → B) → D),
we wish to construct a selection functional
εQi Ki ∈ ((Dω → B) → Dω ).
The idea, which iterates the proof of Proposition 4.4, is to let
εQi Ki (p) = x0 x1 x2 . . . xn . . . ,
where
Q
x0 ∈ K0 is such that ∃α ∈ Qi Ki+1 .p(x0 α),
x1 ∈ K1 is such that ∃α ∈ i Ki+2 .p(x0 x1 α),
...
Q
xn ∈ Kn is such that ∃α ∈ i Ki+n+1 .p(x0 x1 . . . xn α),
...
12
ESCARDÓ
The component xn will be found using εKn , and existential quantifications will be recursively
reduced to search. To make this precise, we change notation. Given a sequence
ε ∈ ((D → B) → D)ω ,
such that εi is a selection functional for Ki , we wish to find
Π(ε) ∈ (Dω → B) → Dω
Q
that is a selection functional for i Ki . That is, we are looking for a computable functional
Π : ((D → B) → D)ω → ((Dω → B) → Dω )
that transforms any sequence of selection functionals for subsets of D into a selection functional for a subset of Dω :
Y
Π(ε)(p)(n) = xn such that ∃α ∈
Ki+n+1 .p(x0 x1 . . . xn α).
i
To complete the derivation of the functional Π, we reduce the existential quantification to
a suitable recursive call to Π. If the functional
Q Π is to meet its specification, Π(λi.εi+n+1 )
should be a selection functional for the set i Ki+n+1 . But a searchable set is exhaustible
by Lemma 3.4. To implement the proof of this lemma in our situation, for any given p, n, xn ,
define
pn,xn (α) = p(x0 x1 . . . xn−1 xn α)
= p(Π(ε)(p)(0) ∗ Π(ε)(p)(1) ∗ . . . ∗Π(ε)(p)(n − 1) ∗ xn ∗ α).
For the sake of clarity, here we have used “∗”, rather than juxtaposition as above, to indicate
concatenation of elements and sequences. Then
Y
∃α ∈
Ki+n+1 .p(x0 x1 . . . xn α)
i
is equivalent to
pn,xn (Π(λi.εn+i+1 )(pn,xn )).
To find xn such that this holds, we use εn :
Π(ε)(p)(n) = εn (λxn .pn,xn (Π(λi.εn+i+1 )(pn,xn ))).
Because we don’t want a different variable xn for each n, we rename the variable to simply x.
This completes our derivation of the product functional:
Definition 4.5. The product functional Π : ((D → B) → D)ω → ((Dω → B) → Dω ) is
recursively defined by
Π(ε)(p)(n) = εn (λx.pn,x,ε (Π(ε(n+1) )(pn,x,ε )))
where



Π(ε)(p)(i) if i < n,


pn,x,ε (α) = p λi. x
if i = n, 


αi−n−1
if i > n,

and where for any sequence β we write β (k) to denote the sequence β with the first k
elements removed:
(k)
βi = βk+i .
EXHAUSTIBLE SETS
13
For future use, we also write β 0 = β (1) and
β(n) = hβ0 , ..., βn−1 i
so that
pn,x,ε (α) = p(Π(ε)(p)(n) ∗ x ∗ α).
The original proof of the following theorem, sketched in [19], uses an auxiliary recurrence
relation and dependent choices. The following more elegant proof, based on alternative
recurrences and bar induction, was presented to me by Ulrich Berger and is included with
his permission:
Theorem 4.6. If eachQεi is a selection functional for a set Ki ⊆ D then Π(ε) is a selection
functional for the set i Ki ⊆ Dω .
Proof. Define
px (α) = p(x ∗ α),
xε,p = Π(ε)(p)(0) = ε0 (λx.px (Π(ε0 )(px ))),
pε = pxε,p = p0,xε,p ,ε .
Claim:
(1)(n) pn+1,x,ε = (pε )n,x,ε0 ,
(2)(n) Π(ε)(p)(n + 1) = Π(ε0 )(pε )(n),
(3)
Π(ε)(p) = xε,p ∗ Π(ε0 )(pε ).
We prove the properties (1)(n) and (2)(n) simultaneously by course of values induction.
Proof of (1)(n) assuming (2)(k) for all k < n: By the assumption,
Π(ε)(p)(n + 1) = Π(ε)(p)(0) ∗ hΠ(ε)(p)(1), ..., Π(ε)(p)(n)i
= xε,p ∗ hΠ(ε0 )(pε )(0), ..., Π(ε0 )(pε )(n − 1)i
= xε,p ∗ Π(ε0 )(pε )(n).
Hence
pn+1,x,ε (α) = p(Π(ε)(p)(n + 1) ∗ x ∗ α)
= p(xε,p ∗ Π(ε0 )(pε )(n) ∗ x ∗ α)
= pε (Π(ε0 )(pε )(n) ∗ x ∗ α)
= (pε )n,x,ε0 (α).
Proof of (2)(n) assuming (1)(n):
Π(ε)(p)(n + 1) = εn+1 (λx.pn+1,x,ε (Π(ε(n+2) )(pn+1,x,ε )))
= ε0n (λx.(pε )n,x,ε0 (Π(ε0(n+1 )((pε )n,x,ε )))
= Π(ε0 )(pε )(n).
Proof of (3): We have (Π(ε)(p))0 = Π(ε0 )(pε ) by (2), and hence
Π(ε)(p) = Π(ε)(p)(0) ∗ (Π(ε)(p))0 = xε,p ∗ Π(ε0 )(pε ).
This completes the proof of the claim.
For a subset L of a domain E and q : E → B defined on L, say that an element x ∈ L
solves q over L if q(x) 6= ⊥, and q(x) = 1 provided q(y) = 1 for some y ∈ L. Then
14
ESCARDÓ
δ : (E → B) → E is a selection function
for L iff δ(q) solves q over L for every q defined
Q
on L. Define the proposition S(p, i Ki ) by
Y
Q
S(p,
Ki ) ⇐⇒ Π(ε)(p) solves p over i Ki whenever εi is a
i
selection functional for Ki .
Q
Q
We need to show that if Q
p is defined on i Ki then S(p, i Ki ) holds. The set of continuous
predicates p defined on i Ki can be defined as follows by bar induction:
Q
(1) if p(⊥) 6= ⊥ then pQis defined on i Ki , and
Q
(2) if px is defined on i Ki+1 for all x ∈ K0 then p is defined on i Ki .
Therefore, it suffices to show that for all p ∈ (Dω → B),
Q
(i) if p(⊥) 6=
Q ⊥ then S(p, i Ki ), and
Q
(ii) if S(px , i Ki+1 ) for all x ∈ K0 then S(p, i Ki ).
Proof of (i): Let b = p(⊥) and assume that b 6= ⊥. Then p = λα.b, by monotonicity
of p. Let εi be a selection function for Ki . Then Π(ε)(p)(n)Q= εn (λx.b). Since λx.b is
defined
Q on Kn , it follows that εn (λx.b) ∈ Kn . Hence Π(ε)(p) ∈ i Ki . If p(α) = 1 for some
α ∈ i Ki then p = λα.1 and hence p(Π(ε)(p)) = 1.
Proof of (ii): Assume the bar induction hypothesis
Q
(†) S(px , i Ki+1 ) for all x ∈ K0 .
We need to show that if εi is a selection function for Ki then:
Q
(ii)(a) Π(ε)(p) ∈ i Ki .
Q
(ii)(b) If p(α) = 1 for some α ∈ i Ki , then p(Π(ε)(p)) = 1.
Proof of (ii)(a): We show that Π(ε)(p)(n) ∈ Kn by induction on n.
Base case for (ii)(a): Π(ε)(p)(0) = xε,p = ε0 (λx.px (Π(ε0 )(px ))). QSince e0i is a selection
function for Ki+1 , it follows from (†) that Π(ε0 )(px ) solves px over i Ki+1 for all x ∈ K0 .
Then λx.px (Π(ε0 )(px )) is defined on K0 , and since ε0 is a selection function for K0 , it follows
that Π(ε)(p)(0) ∈ K0 .
Induction step for (ii)(a): Π(ε)(p)(n + 1) = Π(ε0 )(pεn ) = Π(ε0 )(pxε,p )(n) ∈ Kn+1 by
Claim (2), by (†) and by the fact that xε,p = Π(ε)(p)(0) ∈ K0 (base case).
Q
Q
0
0
Proof of (ii)(b): Assume p(α)
Q and pα0 (α ) = 1,
Q = 1 for some α 0∈ i Ki . Then α ∈ i Ki+1
and, by (†), we have
QS(pα0 , i Ki+1 ). Since ε is a selection function for i Ki+1 , it follows
that Π(ε0 )(pα0 ) ∈ i Ki+1 and pα0 (Π(ε0 )(pα0 )) = 1. Then pε (Π(ε0 )(pε )) = 1 because ε0
is a selection function for K0 . But pε (Π(ε0 )(pε )) = p(xε,p ∗ Π(ε0 )(pε )) = p(Π(ε)(p)), by
Claim (3).
Examples 4.7.
(1) The Cantor space 2ω ⊆ Bω is searchable. A selection functional is given by Π(λi.ε2 )
where ε2 is a selection functional for the finite set 2 ⊆ B.
(2) If Ki Q
⊆ N is a sequence of finite sets that are finitely enumerable uniformly in i,
then i Ki ⊆ N ω is searchable, again using the product functional.
Q
If a product i Ki is searchable, then eachQset Kn is searchable uniformly in n, by Proposition 4.3 as it is the computable image of i Ki under the n-th projection.
EXHAUSTIBLE SETS
15
Remark 4.8. Berger’s selection algorithm ε : (B ω → B) → B ω for the Cantor space, mentioned in the introduction, can be written as
(
0 ∗ ε(λα.p(0 ∗ α)) if p(0 ∗ ε(λα.p(0 ∗ α))),
ε(p) =
1 ∗ ε(λα.p(1 ∗ α)) otherwise.
If one defines ∃ : (B ω → B) → B by ∃(p) = p(ε(p)), as in the proof of Lemma 3.4, then the
above definition is equivalent to
(
0 ∗ ε(λα.p(0 ∗ α)) if ∃α.p(0 ∗ α),
ε(p) =
1 ∗ ε(λα.p(1 ∗ α)) otherwise.
Our product algorithm is inspired by this idea.
From now on, we rely on Section 2.2 for the definition of totality. By Lemma 3.4
above and by Theorem 6.3 below, a non-empty set of total elements is exhaustible iff
it is searchable, and hence the above theorem shows that non-empty, exhaustible sets of
total elements are closed under countable products. For Sierpinski-valued, rather than
boolean-valued, universal quantification functionals, a countable-product algorithm is given
in [18], but we don’t know how to approach countable products of boolean-valued quantifiers
without the detour via selection functionals at the time of writing.
We now derive a uniform continuity principle from Theorem 4.6, motivated by topological theorems that assert that, in certain contexts, continuous functions are uniformly
continuous on compact sets. Define
(
αi i < n,
α =n β ⇐⇒ αi = βi for all i < n,
α|n (i) =
⊥ otherwise.
Then α =n β iff α|n = β|n .
Theorem 4.9. If f ∈ (Dω → N ) is definedQon a product
there is a number n such that for all α, α0 ∈ i Ki ,
Q
i Ki
of searchable sets, then
α =n α0 =⇒ f (α) = f (α0 ).
Proof. Let (==) ∈ (N × N → B) be the unique total function such that (x == y) = 1 iff
x ∼ y. Then ∀Qi Ki (λα.f (α) == f (α)) = 1. If we define
f|n (α) = f (α|n ),
F
then f = n f|n and hence (λα.f (α) == f (α)) = n (λα.f|n (α) == f (α)). So, by continuity of ∀Qi Ki , there is n such that
F
∀Qi Ki (λα.f|n (α) == f (α)) = 1.
Q
We cannot conclude that f|n (α) == f (α) for all α ∈ i Ki because
there is no reason
Q
why the predicate Q
λα.f|n (α) == f (α) should be defined on i Ki . To overcome this
difficulty,
let β ∈ i Ki+n and define gn (α) = f (α0 α1 . . . αn−1 β) so that gn is defined
Q
on i Ki and above fn . By monotonicity, ∀Qi Ki (λα.gn (α) == f (α)) = 1. Now the predicate
Q
Q
λα.gn (α) == f (α) is defined on i Ki and hence gn (α) = f (α) for all α ∈ i Ki . But if
α =n α0 then gn (α) = gn (α0 ), and so f (α) = f (α0 ), as required.
16
ESCARDÓ
The following is an immediate consequence of this and Theorem 4.6:
Corollary 4.10. The functional fan = fanQi Ki : (Dω → N ) → N defined by
Y
fan(f ) = µn.∀α, β ∈
Ki . α =n β =⇒ f (α) = f (β)
i
is computable uniformly in any sequence
of selection functionals for the sets Ki ⊆ D, and is
Q
defined on any f that is defined on i Ki . Moreover, if the sets Ki consist of total elements
of a domain D = Dσ , then the fan functional is total.
This holds, in particular, if D = N and each Ki is a finite subset of N defined uniformly
in i, which is the case that has been considered in higher-type computability theory regarding the fan functional (see e.g. [23]). Here we have generalized this to arbitrary higher types
D = Dσ . A consequence of the exhaustibility of the Cantor space is that:
Corollary 4.11. The total elements of the function space (B ω → N ) have decidable equivalence.
Proof. The algorithm (==) : (B ω → N ) × (B ω → N ) → B given by
(f == g) = ∀α ∈ 2ω .f (α) == g(α)
does the job.
This can be generalized as follows, where we now rely on Section 2.3 for the definition
of the Kleene–Kreisel spaces Cσ .
Definition 4.12. The discrete and compact types are inductively defined as
discrete ::= o | ι | discrete × discrete | compact → discrete,
compact ::= o | compact × compact | discrete → compact.
The reason for this terminology is that the space Cσ is discrete if σ is discrete, and it is
compact if σ is compact, as observed in [18].
Theorem 4.13.
(1) If σ is discrete, then Cσ is computably enumerable.
(2) The total elements of a domain of compact type form a searchable set.
(3) The total elements of a domain of discrete type have decidable equivalence.
Proof. By induction on the definitions of discrete and compact type. The first condition
holds by the Kleene–Kreisel density theorem, which gives a computable dense sequence
of Cσ , and by the fact that Cσ is discrete. For the second condition, use Theorem 4.6 with
the aid of the first condition, and, for the third one, use the argument of Corollary 4.11.
We conclude this section with a natural notion that plays a fundamental role in our
investigation of exhaustible and searchable sets and their relationship. Let D = Dσ and
D0 = Dσ0 for types σ and σ 0 .
Definition 4.14. We say that a set K ⊆ D is entire if it consists of total elements and is
closed under total equivalence.
EXHAUSTIBLE SETS
17
Notice that if p is total then it is defined on every entire set. If p is not total and K
is not entire, but if p is defined on K, then p(x) = p(x0 ) for all x ∼ x0 in K, because if
x ∼ x0 then x and x0 are bounded above and hence so are p(x) and p(x0 ), which then must
be equal as they are non-bottom by definition. But if x ∈ K and x0 ∼ x for x0 outside K,
it doesn’t follow that p(x0 ) 6= ⊥ (consider e.g. K = {λi.1} for σ = ι → o and p(α) = α(⊥)).
The following closure properties of entire sets are easily verified:
(1) If K ⊆ D and K 0 ⊆ D0 are entire, so is K × K 0 Q
⊆ D × D0
(2) If Ki is a sequence of entire subsets of D, then i Ki is an entire subset of Dω .
Definition 4.15. The image of an entire set by a total function doesn’t need to be entire,
but it consists of total elements, and hence its closure under total equivalence is entire. We
refer to this as its entire image. (Thus, entire images are defined for total functions and
entire sets only.)
Proposition 4.16. Exhaustible and searchable sets are closed under the formation of computable entire images.
Proof. For given f : D → D0 and K ⊆ D exhaustible, consider the quantification functional
∀f (K) (q) = ∀x ∈ K.q(f (x)).
defined in the proof of Proposition 4.3. If f is total and K is entire with entire image L,
then we can take ∀L = ∀f (K) . To verify this, let q be defined on L. Then q is defined on
f (K) ⊆ L, and hence if q(l) = 1 for all l ∈ L, then ∀L (q) = 1. If, on the other hand, q(l) = 0
for some l ∈ L, then l ∼ f (x) for some x ∈ K. But then q(f (x)) = 0, and so ∀T (q) = 0,
which concludes the verification. The argument for searchable sets is similar.
Definition 4.17. Let S = {⊥, >} by the Sierpinski domain and F ⊆ D = Dσ be entire.
(1) F is decidable if there is a total computable map ψF : D → B such that, for all total
x ∈ D, ψF (x) = 1 iff x ∈ F .
(2) F is semi-decidable if there is a computable map χF : D → S such that, for all total
x ∈ D, ψF (x) = > iff x ∈ F .
(3) F is co-semi-decidable if its complement in T = Tσ semi-decidable.
Notice that the functions ψF is not uniquely determined by F , because its behaviour
is specified on a subset of D, but that χF is uniquely determined by F . Notice also F is
decidable if and only if it is decidable on T in the sense of Definition 4.1 with K = T .
5. Compactness of exhaustible sets
A notion analogous to exhaustibility, with the Sierpinski domain S playing the role of
the boolean domain B, is considered in [18]. A crucial fact, formulated here as Lemma 2.2,
is that the (now unique) quantification functional ∀K : (D → S) → S is continuous iff
the set K is compact in the Scott topology of D. Hence, because computable functionals
are continuous, Sierpinski-exhaustible sets are compact, and so Sierpinski exhaustibility is
seen as articulating an algorithmic version of the topological notion of compactness. The
computational idea is that, given any semi-decidable property of D, one can semi-decide
whether it holds for all elements of K. Closure properties analogous to the above are
established for Sierpinski exhaustibility in [18].
The present investigation can be seen as a natural follow-up of that work that arises
by asking what changes if one moves from semi-decision problems to decision problems.
18
ESCARDÓ
One significant change is that continuity of a quantification functional ∀K : (D → B) → B
doesn’t entail the compactness of K in the Scott topology any longer:
Examples 5.1.
(1) There are exhaustible sets that fail to be compact in the Scott topology.
By [40, 36], any second-countable T0 space, e.g. the real line R, can be embedded
into the domain D = B ω under the Scott topology. But R is a connected space,
which is equivalent to saying that every continuous boolean-valued map defined on
it is constant. Hence a predicate p ∈ (D → B) is defined on R iff it is constant on R.
Therefore R is trivially exhaustible: ∀R (p) = p(0). But it is not compact.
Notice also that any space embedded into the total elements of B ω must be totally
disconnected, and hence any embedding of R into B ω must assign non-total elements
of B ω to some real numbers. One may suspect that if such embeddings are ruled
out, this problem would disappear. But this is not the case, as the next example
shows.
(2) There are exhaustible sets of total elements that fail to be Scott compact.
In fact, there is a trivial and pervasive counter-example. Let f ∈ ((N → N ) → N )
be total. Then the total equivalence class K of f , as is well known and easy to
verify, doesn’t have minimal elements, and hence cannot be compact in the Scott
topology. But it is exhaustible with ∀K (p) = p(f ).
One may feel somewhat cheated by the second counter-example, because although the
set K is not Scott compact, it is generated by the singleton {f }, which is Scott compact,
and because we took ∀K to be ∀{f } (cf. the proof of Proposition 4.16). Lemma 5.5(3) below
shows that any counter-example is generated by a Scott compact set in a similar fashion.
In any case, although exhaustible sets do fail to be compact in the Scott topology, if they
consist of total elements then they are compact in the Kleene–Kreisel topology. In order to
formulate and prove this, we need some definitions. We now rely on Section 2.3.
Definition 5.2. Let σ be a type, D = Dσ , T = Tσ , C = Cσ and ρ = ρσ : Tσ → Cσ .
(1) By the shadow of a set K ⊆ T we mean its ρ-image in C. Similarly, by the shadow
of an element x ∈ T we mean its ρ-image ρ(x) in C.
(2) A set K ⊆ T is called Kleene–Kreisel compact if its shadow is compact.
Recall that the Cantor space is the set 2ω of maximal elements of B ω .
Remark 5.3. Sometimes, for example for the implementation of the product functional
defined in Section 4 in the language PCF, which lacks countable powers, one works with
the Cantor space within the function space B N . The Cantor space is homeomorphic to
the subspace of total strict functions α ∈ BN , where α is strict if α(⊥) = ⊥. It is also
homeomorphic to the quotient of the set of all total elements of B N . But notice that the
set of maximal elements of B N is not homeomorphic to the Cantor space. This is because
the two non-strict elements λi.0 and λi.1 are finite (or order compact), and hence isolated in
the relative Scott topology (meaning that the two corresponding singletons are open), and
hence the maximal elements have a topology strictly finer than that of the Cantor space, as
there are no isolated points in the Cantor space. As is well known in topology, no compact
Hausdorff topology can have another compact Hausdorff topology as a strict refinement.
EXHAUSTIBLE SETS
19
Every (computationally) exhaustible set is topologically exhaustible in the sense of the
following definition, because computable maps are continuous.
Definition 5.4. We say that a set K ⊆ D is topologically exhaustible if there is a continuous
map ∀K ∈ ((D → B) → B) satisfying the conditions of Definition 3.2.
The following is our main tool in the constructions and proofs of correctness and termination
of algorithms developed in Sections 6–7. Its proof relies on Sections 2.4 and 2.5.
Lemma 5.5.
(1) Any topologically exhaustible set of total elements is Kleene–Kreisel compact.
(2) Any non-empty, Kleene–Kreisel compact entire set is an entire continuous image of
the Cantor space and hence is topologically exhaustible.
(3) Any Kleene–Kreisel compact entire set has a Scott compact subset with the same
shadow.
Proof. (1): Let K ⊆ T be exhaustible. By Lemma 2.6 and the fact that clopen sets are
closed under finite unions, to establish compactness of ρ(K), it is enough to consider a
directed clopen cover U. By Lemma 2.5, for every U ∈ U there is a total pU ∈ (D → B)
with
−1
−1
(†) ρ−1 (U ) ⊆ p−1
U (1) and ρ (C \ U ) ⊆ pU (0).
Define predicates qU , r ∈ (D → B) by
[
−1
qU−1 (1) = p−1
(1),
r
(1)
=
p−1
qU−1 (0) = r−1 (0) = ∅.
U
U (1),
U ∈U
F
S
Then qU v pU , the set {qU | U ∈ U} is directed, and r = U ∈U qU . Because ρ(K) ⊆ U,
we have that K ⊆ r−1 (1) and hence ∀K (r) = 1. So, by continuity of ∀K , there is U ∈ U with
∀K (qU ) = 1, and hence with ∀K (pU ) = 1 by monotonicity. Let x ∈ K. Then pU (x) = 1
by specification of ∀K and the fact that pU is total and hence defined on K. But then
ρ(x) ∈ U , for otherwise (†) would entail pU (x) = 0. This shows that ρ(K) ⊆ U , and so
ρ(K) is compact.
(2): By e.g. [21], any compact subset of C is countably based (even though C is not).
But any non-empty compact Hausdorff countably based space is a continuous image of the
Cantor space. Hence there is a continuous map 2ω → C with image ρ(K) for any entire set
K ⊆ D. Then the entire image of the Cantor space under any representative B ω → D is K.
(3): This follows from the argument given in (2), because the Cantor space is Scott
compact.
Remark 5.6. In particular, this gives a topological view of the computational fact stated
in the introduction that exhaustible sets of natural numbers must be finite: all compact
sets are finite in a discrete space.
Kleene–Kreisel compactness can be expressed as a finite-subcover condition for the Scott
topology as follows: An entire set K ⊆ D is Kleene–Kreisel compact if and only if every
cover of K by Scott open sets that are closed under total equivalence has a finite subcover.
This is a straightforward consequence of the fact that the Kleene–Kreisel topology is the
quotient by total equivalence of the relative Scott topology on the total elements.
We also remark that there is a natural topology on D, coarser than the Scott topology,
in which all exhaustible sets are compact. Part of the argument of Lemma 5.5(1) shows
20
ESCARDÓ
that any exhaustible set K is compact in the coarsest topology on D such that all predicates
p ∈ (D ∈ B) defined on K are continuous. This is generated by directed unions of basic open
sets of the form p−1 (1) with p as above, because such sets are closed under finite unions
and intersections. This construction is analogous to the zero-dimensional reflection of a
topology, and happens to coincide with it in the case considered in Lemma 5.5(1), modulo
quotienting, and can also be compared with the weak topology in functional analysis.
6. Searchability of exhaustible sets
We already know that every searchable set is exhaustible (Lemma 3.4). This implication
is uniform, in the sense that there is a computable functional
((D → B) → D) → ((D → B) → B)
that transforms selection functionals into quantification functionals, namely ε 7→ λp.p(ε(p)).
We now establish the converse for non-empty entire sets, and some additional results. The
fact that exhaustible entire sets are Kleene–Kreisel compact, established in the previous
section, plays a fundamental role in the construction of the algorithms
Definition 6.1. We say that a set S ⊆ D = Dσ is a retract up to total equivalence if there
is a function r ∈ (D → D) such that
(1) r(x) ∈ S for all total x ∈ D,
(2) r(s) ∼ s for all s ∈ S.
In this case, r is total, all elements of S are total, and r(r(x)) ∼ r(x) for all total x.
Notice that r is a retract up to total equivalence iff it is a total function and its Kleene–
Kreisel shadow ρ(r) : C → C is a retract in the usual topological sense, where C = Cσ .
Definition 6.2. We say that two entire sets K ⊆ D = Dσ and L ⊆ E = Dτ are homeomorphic up to total equivalence if there are total functions f ∈ (D → E) and g ∈ (E → D)
such that g(f (x)) ∼ x and f (g(y)) ∼ y for all x ∈ K and y ∈ L.
This is equivalent to saying that the shadow functions ρ(f ) and ρ(g) restrict to a (true)
homeomorphism between the shadows of K and L. In this case, L is the entire f -image
of K, and K is the entire g-image of L.
Theorem 6.3. If K ⊆ D = Dσ is a non-empty, exhaustible entire set then, uniformly in
any quantification functional for K:
(1) K is searchable.
(2) K is a computable entire image of the Cantor space.
(3) K is computably homeomorphic to some entire exhaustible subset of the Baire domain N ω , up to total equivalence.
(4) K is a computable retract up to total equivalence.
(5) K is co-semi-decidable.
In particular, after the theorem is proved, one can w.l.o.g. work with total predicates
rather than predicates defined on K, as for any predicate p ∈ (Dσ → B) defined on K
one can uniformly find a total predicate that agrees with p on K, by composition with the
retraction.
EXHAUSTIBLE SETS
21
Proof. We proceed by cases, of increasing generality, on the type of K. The case K ⊆ N
is trivial and is implicitly used in the case K ⊆ N ω , which in turn is used in the next case
K ⊆ (D → N ). The general case K ⊆ D is reduced to this last case via retracts using
Lemma 2.1.
(i) Case K ⊆ N : We can define
(
µn.∃m ∈ K.n = m ∧ p(n) if ∃n ∈ K.p(n),
εK (p) =
µn.∃m ∈ K.n = m
otherwise.
This construction defines εK uniformly in ∃K . We could now easily show that K satisfies
the other conditions of the theorem, but this won’t be required for our proof, as this will
follow in later cases.
For future use, notice that if K ⊆ N is entire and exhaustible, then the supremum of
the finite set K (which is zero if K is empty and the largest element of K otherwise) can
be computed uniformly in any quantification functional for K as
sup K = µm.∀n ∈ K.n ≤ m.
Hence, the finite enumeration en of the elements of K, in ascending order, for 0 ≤ n <
cardinality(K), is uniformly computable as
en = µy.∃m ∈ K.∀i < n.m 6= ei ∧ m = y.
We stop when we find n such that en = sup K, and we include en iff ∃m ∈ K.m = sup K.
(ii) Case K ⊆ N ω . We first argue that we can find some α ∈ K, uniformly in ∃K , by
the following algorithm defined by course-of-values induction on n:
αn = µk.∃β ∈ K.α =n β ∧ βn = k.
Recall that we defined
α =n β ⇐⇒ ∀i < n.βi = αi
in the paragraph preceding Theorem 4.9. By construction, for every n there is β ∈ K with
α =n β, and in particular α is total. Because the shadow of K is compact, it is closed, and
because K is entire, α ∈ K, as required. Then we can define, using Proposition 4.2 and the
above algorithm to construct α in both cases,
(
some α ∈ K ∩ p−1 (1) if ∃α ∈ K.p(α),
εK (p) =
some α ∈ K
otherwise.
Again, this construction defines εK uniformly in ∃K .
We now show that K is a computable retract up to total equivalence, uniformly in any
quantification functional for K. Define
r = rK : N ω → N ω
by course-of-values induction on n:
(
αn
r(α)(n) =
µm.∃β ∈ K.β =n r(α) ∧ βn = m
if ∃β ∈ K.β =n r(α) ∧ βn = αn ,
otherwise.
Because the shadow of K is closed, the finite prefixes of its members form a tree whose
infinite paths correspond to the elements of K. The above algorithm follows the infinite
path α through the tree, either for ever (always following the first case) or until the path
22
ESCARDÓ
exits the tree (reaching the second case). If and when α exits the tree, we replace the
remainder of α by the left-most infinite branch of the subtree at which α exits the tree.
Then r clearly satisfies the required conditions.
A semi-decision procedure for the complement of K is given by
α 6∈ K ⇐⇒ r(α) 6= α,
using the fact that apartness of total elements of N ω is semi-decidable. (This is a computational version of the topological fact that retracts of Hausdorff spaces are closed.)
We now show that K is a computable entire image of the Cantor space. For any i, the
set Ki = {αi | α ∈ K} is Q
exhaustible by Proposition 4.3 as evaluation at i is computable.
It is enough to show that i {0, 1, . . . , sup Ki } ⊆ N ω is an entire image of the Cantor space
by a computable map t : B ω → N ω , because then r ◦ t has K as its entire image since K is
contained in that product. But this is straightforward: at each stage j of the computation
of t(α), look at the next dlog2 (sup Kj )e digits of the input α, compute the natural number
f (j) represented by this finite sequence, and let t(α)(j) = min(sup Kj , f (j)).
(iii) Case K ⊆ (D → N ) where D = Dσ for an arbitrary type σ: In order to reduce this
to case (ii), we invoke the Kleene–Kreisel density theorem, to get a computable sequence
d ∈ Dω such that the shadow sequence ρ(dn ) is dense in C = Cσ . Define
P : (D → N ) → N ω
P (f ) = λn.f (dn ).
We will define a total function E = EK in the other direction,
E : N ω → (D → N )
such that
R = E ◦ P : (D → N ) → (D → N )
exhibits K as a retract of (D → N ) up to total equivalence. Thus, for f ∈ K, one can
recover the behaviour of f at total elements from its behaviour on the dense sequence d.
Because this implies that K is the entire image of P (K), and because P (K) is searchable
by case (ii), it will follow that K is searchable and an entire image of the Cantor space,
because E preserves total equivalence on P (K).
For α ∈ N ω and n ∈ N, define
Fnα = {f ∈ (D → N ) | ∀i < n.f (di ) = αi },
Knα = K ∩ Fnα .
Here we regard α as potentially coding the action of some f on the set of elements di . The
set Fnα is decidable on K uniformly in α and n, and hence Knα is uniformly exhaustible by
Proposition 4.2.
Lemma 6.4. If K ⊆ (D → N ) is a Kleene–Kreisel compact entire set, then for all total
α ∈ N ω and all total x ∈ D there is n such that f (x) = f 0 (x) for all f, f 0 ∈ Knα .
Proof. For any g ∈ Cσ→ι the set BgT= {f ∈ Cσ→ι | f (x) = g(x)} is clopen, where we write
α
α
x = ρ(x). By density of dn , the
T set α n K nThas atα most one element, where K n denotes the
α
shadow of Kn . Hence if g ∈ n K n then n K n = {g} ⊆ Bg . Because Cσ→ι is Hausdorff
and because each K αn is compact and Bg is open, there is n such that already K αn ⊆ Bg . So
for all f ∈ K αn one has f (x) = g(x), and hence for all f , f 0 ∈ K αn one has f (x) = f 0 (x).
EXHAUSTIBLE SETS
23
By Proposition 4.16, the entire P -image L ⊆ N ω of K is exhaustible. Let r = rL be
defined as in case (ii), and define E : N ω → (D → N ) by
E(α)(x) = µy.∃f ∈ Knr(α) .f (x) = y,
r(α)
where n is the least number such that ∀f, f 0 ∈ Kn .f (x) = f 0 (x). By exhaustibility
r(α)
of Kn , this can be found uniformly in α, and hence E is computable uniformly in K.
Proof of correctness of E.
(a) E is total and maps L into K. Let α ∈ N ω be total. Then r(α) ∈ L, by construction
r(α)
of r, and hence there is g ∈ K with r(α) ∼ P (g), and so with g ∈ Kn
for any n. Let
r(α)
0
0
x ∈ D be total and n be the least number such that f (x) = f (x) for all f, f ∈ Kn . Then
r(α)
f (x) = g(x) for all f ∈ Kn , and hence E(α)(x) = g(x). Therefore E(α) ∼ g ∈ K, and
hence E(α) ∈ K as K is entire, and in particular E is total. By construction E ∼ E ◦ r,
and hence, because r exhibits L as a retract up to total equivalence and K is entire, the
E-image of L is K.
(b) If f ∈ (D → N ) is total then R(f ) = E(P (f )) ∈ K. Because P (f ) ∈ L.
(c) If f ∈ K then R(f ) ∼ f . Continuing from the proof of (a), for α = P (f ) we have
r(α) ∼ α by construction of r, and hence for any g ∈ K such that P (g) = r(α) we have
g(di ) = P (g)(i) = r(α)(i) = αi = P (f )(i) = f (di ) and so g ∼ f by density, which shows
that R(f ) = E(P (f )) ∼ f , as required.
A semi-decision procedure for the complement of K is given as in case (ii),
f 6∈ K ⇐⇒ R(f ) 6= f,
f0
because f 6=
⇐⇒ ∃n ∈ N.f (dn ) 6= f 0 (dn ), for total functions f 0 and f since K is entire
and d is dense.
Because E and P are total, they induce computable Kleene–Kreisel functionals E =
ρ(E) : NN → NC and P = ρ(P ) : NC → NN where C = Cσ . If K ⊆ NC is the shadow of K,
then the restriction of P to K followed by the co-restriction to its image is a homeomorphism
K → P (K): abstractly because any continuous bijection of compact Hausdorff spaces is
a homeomorphism, and concretely because the bi-restriction of E is a continuous inverse.
Hence any exhaustible subset of (D → N ) is computably homeomorphic to the shadow of
some exhaustible subset of the Baire domain N ω , up to total equivalence.
(iv) General case. We derive this from the case (iii). By Lemma 2.1, for any D = Dσ
there are D0 = Dτ →ι and computable P : D0 → D and E : D → D0 such R = E ◦ P is
a retraction up to total equivalence and Tσ is the entire image of P . Let K ⊆ D be a
non-empty, exhaustible entire set, and let K 0 be the entire E-image of K. Then K is the
entire image of P (K 0 ), and, because K is entire, a predicate p0 ∈ (D0 → B) defined on K 0
holds for all x0 ∈ K 0 if and only if p0 ◦ E holds for all x ∈ K. Hence K 0 is exhaustible with
∀K 0 (p0 ) = ∀K (p0 ◦ E). By case (iii) above, K 0 is searchable. Therefore K is searchable by
Proposition 4.3. Similarly, the other properties we need to establish are closed under the
formation of retracts and hence are inherited from case (iii).
This concludes the proof of Theorem 6.3.
24
ESCARDÓ
7. Ascoli–Arzela type characterizations of exhaustible sets
We reformulate a theorem of Gale’s [22] that characterizes compact subsets of function
spaces (Theorem 7.1). This suggests a characterization of exhaustible entire sets (Theorem 7.5), whose topological version is developed first (Theorem 7.4). The main idea is to
replace a condition in Gale’s theorem by a continuity condition (Section 7.1), and then
further replace it by a computability condition (Section 7.2). This method of transforming topological theorems into computational theorems is the main thrust of the paper [18],
which develops many instances of computational manifestations of topological theorems.
7.1. Topological version. The Heine–Borel theorem characterizes the compact subsets
of Euclidean space Rn as those that are closed and bounded. The Arzela–Ascoli theorem
generalizes this to subsets of RX , where X is a compact metric space and RX is the set of
continuous functions endowed with the metric defined by
d(f, g) = max{d(f (x), g(x)) | x ∈ X}.
A set K ⊆ RX is compact if and only if it is closed, bounded and equi-continuous. Equicontinuity of K means that the functions f ∈ K are simultaneously continuous, in the
sense that for every x ∈ X and every > 0, there is δ > 0 such that d(x, x0 ) < δ =⇒
d(f (x), f (x0 )) for all x0 ∈ X and all f ∈ K. The Heine–Borel theorem is the particular case
in which X is the discrete space {1, . . . , n}, for equi-continuity holds automatically for any
subset of RX in this case. The above metric on RX induces the compact-open topology.
More general Arzela–Ascoli type theorems characterize compact subsets of spaces Y X of
continuous functions under the compact-open topology, for a variety of spaces X and Y ,
with a number of generalizations or versions of the notion of equi-continuity, notably even
continuity in the sense of Kelley [28].
Among a multitude of generalizations of the Arzela–Ascoli theorem, that of Gale [22,
Theorem 1] proves to be relevant concerning exhaustibility of entire sets:
If X and Y are Hausdorff k-spaces with Y regular, a set K ⊆ Y X is compact
if and only if
(1) K is closed,
(2) the set K(x)
= {f (x) | f ∈ K} is compact for every x ∈ X,
T
(3) the set f ∈K∩F f −1 (V ) is open for every closed set F ⊆ Y X and for
every open set V ⊆ Y .
Gale didn’t assume Y to be a k-space and formulated this for the compact-open topology, but his theorem holds for the exponential topology if we require Y to be a k-space.
Regarding compactness, we have already mentioned that a Hausdorff space has the same
compact sets as its k-coreflection, and that the exponential topology is the k-reflection of
the compact-open topology. Although there are more closed sets in the exponential topology, Gale’s argument works with closedness of K in the exponential topology. This follows
from the general considerations of Kelley [28, Chapter 7].
The last condition is a version of equi-continuity. Because X is not assumed to be
compact, the set K cannot be globally bounded in any sense, but it is pointwise bounded
in the sense of the second condition. This gives a characterization of compact subsets of
Kleene–Kreisel spaces of the form NC and in particular of Kleene–Kreisel spaces of pure type,
because N is regular. However, as discussed in Section 2.5, Matthias Schröder has recently
EXHAUSTIBLE SETS
25
N
shown that NN is not regular, and this justifies the restriction of our characterizations of
exhaustible entire sets to particular kinds of types in Section 7.2.
Notice that when X = Y = N, this amounts to the well known characterization of
compact subsets K of the Baire space Nω as finitely branching trees. The equi-continuity
condition, as in the case of the Heine–Borel theorem, is superfluous, because any set is equicontinuous in this case as the topology of the exponent is discrete. Condition (1) says that
the elements of K are the paths of a tree, and (2) says that the tree is finitely branching,
because the compact subsets of the base space are finite.
Lemma 2.2 and the remarks preceding it allow one to consider continuity of functions
involving points of a k-space X, open sets and closed sets (using the function space S X and
representing open sets and closed sets by their characteristic functions), and compact sets
X
(using the function space S S and representing compact sets by their universal quantification functionals). We now reformulate Gale’s theorem by expressing condition (3) as a
continuous version of a slight strengthening of condition (2).
Theorem 7.1. If X and Y are Hausdorff k-spaces with Y regular, a set K ⊆ Y X is compact
if and only if
(1) K is closed, and
(2) (K ∩ F )(x) is compact, continuously in F and x, for any closed set F ⊆ Y X and
any x ∈ X.
The dependence of (K ∩ F )(x) in the parameters F and x is given by the functional
Φ : SY
X
× X → SS
Y
defined by Φ(χ̄F , x) = ∀(K∩F )(x) , where we write
χ̄F = χF c .
Proof. (⇒): The set K is closed because Y X is Hausdorff. The set K ∩ F is compact
because F is closed. Because the evaluation map is continuous and because (K ∩ F )(x)
is the continuous image of K ∩ F under evaluation at x, it is compact. To see that Φ is
continuous, let v ∈ S Y . Then v(y) = > for all y ∈ (K ∩ F )(x) ⇐⇒ v(f (x)) = > for all
f ∈ K ∩ F ⇐⇒ f ∈ F c or v(f (x)) for all f ∈ K. Hence
Φ(w, x) = λv.∀f ∈ K.w(f ) ∨ v(f (x)),
where (∨) : S × S → S is defined by a ∨ b = > iff a = > or b = >. Because the functional
∀K is continuous as K is compact, and because the category of k-spaces is cartesian closed
and the above is a λ-definition from continuous maps, Φ is continuous.
(⇐): It suffices to show that Gale’s conditions (1)-(3) hold. Condition (1) is the same
as ours, and Gale (2) follows from our condition (2) with F = Y X . To prove Gale (3), let
F ⊆ Y X be closed and V ⊆ Y be open. Then the set
U = {x ∈ X | Φ(χ̄F , x)(χV ) = >}
is open because Φ is continuous, and
x∈U
⇐⇒
⇐⇒
⇐⇒
∀(K∩F )(x) (v) = > ⇐⇒ χV (y) = > for all y ∈ (K ∩ F )(x)
χV (f (x)) = > for all f ∈ K ∩ F ⇐⇒ f (x) ∈ V for all f ∈ K ∩ F
T
x ∈ f ∈K∩F f −1 (V ),
T
which shows that the set f ∈K∩F f −1 (V ) is the same as U and hence is open.
26
ESCARDÓ
Definition 7.2. We say topologically decidable etc. taking the continuous versions of Definitions 4.1 and 4.17.
We now formulate and prove an analogue of this theorem, which replaces (i) the Sierpinski space S by the boolean domain B, (ii) Hausdorff k-spaces by Scott domains, (iii) compact subsets by topologically exhaustible entire subsets, (iv) closed subsets by topologically
decidable sets (cf. Definitions 3.2 and 5.4). We again apply Gale’s theorem, exploiting
Hyland’s characterization of the Kleene–Kreisel spaces as k-spaces. The proof follows the
same pattern as that of Theorem 7.1, but there are a number of additional steps. Firstly,
using Gale’s theorem, we get continuous maps defined on Kleene–Kreisel spaces. These are
extended to continuous maps on domains using the Kleene–Kreisel density theorem and
Scott’s injectivity theorem, as in Lemma 2.5. (In Theorem 7.5, such an extension will be
instead defined by an algorithm, but still relying on the density theorem.) Secondly, the
set F in condition (2) is closed in Theorem 7.1 but is neither open nor closed in Theorem 7.4, although it has clopen shadow, because the Sierpinski space has been replaced by
the boolean domain. To overcome this difficulty, we rely on the following version of Gale’s
theorem:
Remark 7.3. An inspection of the proof of Gale’s theorem shows that it also holds if, in
condition (3), the set F ranges over subbasic closed sets in the compact-open topology:
T
30 . the set f ∈K∩N (Q,B) f −1 (V ) is open for every compact set Q ⊆ X, every closed set
B ⊆ Y , and every open set V ⊆ Y .
In one direction this is clear: if condition (3) holds for all closed F , then it holds for F =
N (Q, B). For the other direction, notice that condition (3) is used only in the “Lemma”
[22, page 305] for F of this form (the sets Wx in the second last line of that page, and the
set T of page 306).
Let D = Dσ and C = Cσ for an arbitrary type σ, and recall the concepts and notation
introduced in Definitions 4.17 and 5.4.
Theorem 7.4. An entire set K ⊆ (D → N ) is topologically exhaustible if and only if the
following two conditions hold:
1. K is topologically co-semi-decidable.
2. The set (K∩F )(x) is topologically exhaustible for any F that is topologically decidable
on K, and any x ∈ D total, continuously in F and x.
Here the dependence of (K ∩ F )(x) in F and x is to be given by a functional
Γ : ((D → N ) → B) × D → ((N → B) → B)
such that Γ(ψF , x) = ∀(K∩F )(x) .
Proof. (⇒): (1): By Lemma 5.5, the shadow K = ρ(K) ⊆ NC of K is compact and hence
closed. Hence the map NC → B that sends f ∈ K to ⊥ and f 6∈ K to 1 is continuous. By
composition with the quotient map ρ : T → NC , where T = Tσ→ι , we get a map T → B.
Because T is dense in (D → N ) and B is densely injective, the domain Dσ→ι under the
Scott topology is injective over dense embeddings, which means that this map extends to
a continuous map (D → N ) → B. By construction, this exhibits K as a topologically
co-semi-decidable subset of (D → N ).
EXHAUSTIBLE SETS
27
(2): Define Γ(ψF , x) = λp.∀f ∈ K.ψF (f ) =⇒ p(f (x)). The result then follows from
the fact that the category of Scott domains under the Scott topology is cartesian closed,
and hence functions that are λ-definable from continuous maps are themselves continuous.
(⇐): We apply Gale’s theorem to show that the shadow K = ρ(K) is compact. Then
it is topologically exhaustible by Lemma 5.5.
Gale (1): If K is topologically co-semi-decidable, then, by definition, we have a continuous function (D → N ) → B that maps f ∈ K to ⊥ and f 6∈ K to 1. Hence K is closed
in T because it is the inverse image of the closed set {⊥} restricted to T . Because K is
entire, it is closed under total equivalence by definition, and hence, because ρ : T → NC is
a quotient map, K is closed.
Gale (2): The assumption gives that for any x ∈ D total, K(x) is exhaustible, considering F = (D → N ). Because K is entire and x is total, K(x) ⊆ N. Hence by Lemma 5.5,
K(x) is compact in N ⊆ N .
Gale (3): Let F ⊆ NC be a subbasic open set of the form N (Q, V ) with Q ⊆ C
compact and V ⊆ N (necessarily) clopen. Then the set Q = ρ−1 (Q) is entire and Kleene–
Kreisel compact, and hence, by Lemma 5.5, it is topologically exhaustible. Also, V is
a topologically decidable subset of N . So the predicate p : (D → N ) → B defined by
p(f ) = ∀x ∈ Q.χV (f (x)) is continuous and defined on K, and p = ψF for F = T ∩ p−1 (1).
Now define u : D → B by
u(x) = Γ(ψF , x)(χV ).
Then u is continuous and
u(x) = ∀(K∩F )(x) (χV ) = ∀f ∈ K ∩ F.χV (f (x)).
Hence the
u−1 (1) = {x ∈ T | ∀f ∈ K ∩ F.χV (f (x)) = 1} is open. Therefore its
T set U = −1
shadow f ∈K ∩ F f (V ) is open, because it is closed under total equivalence and because
ρ is a quotient map.
7.2. Computational version. At this stage of our investigation, such a characterization
is available only for certain types, which include pure types, and for entire sets (for the
reasons explained in Section 7.1). Let D = Dσ and C = Cσ for an arbitrary type σ. We
establish the computational version of Theorem 7.4.
Theorem 7.5. An entire set K ⊆ (D → N ) is exhaustible if and only if the following two
conditions hold:
1. K is co-semi-decidable.
2. The set (K ∩ F )(x) is exhaustible for any F decidable on K, and any x ∈ D total,
uniformly in F and x.
Moreover, the equivalence is uniform.
A few remarks are in order before embarking into the proof. The claim holds, with
the same proof, if conditions (1) and (2) are replaced by any of the following conditions,
respectively:
10 . K is topologically co-semi-decidable.
100 . K has closed shadow.
1000 . The shadow of K is closed in the topology of pointwise convergence.
10000 . K has compact shadow.
28
ESCARDÓ
20 . The set (K ∩ Fnα )(x) is exhaustible, uniformly in n ∈ N , α ∈ N ω and x ∈ D total.
Recall (proof of Theorem 6.3) that we defined
Fnα = {f ∈ (D → N ) | ∀i < n.f (di ) = αi }.
In the formulation of the theorem, the fact that conditions (1) and (2) uniformly imply
the exhaustibility of K is in principle given by a computable functional of type
χK c
ψF
x
∀K∩F (x)
∀K
z
z
z
}|
{
}|
{ z}|{
z
}|
{
}|
{
((D → N ) → S) × (((D → N ) → B) × D → ((N → B) → B)) → (((D → N ) → B) → B) .
{z
} |
{z
}
{z
}
|
|
condition 1
condition 2
conclusion
However, the computational information given by condition (1) is not used in the construction of the conclusion (although the topological information is used in its correctness
proof). Moreover, the information given by condition (2) is not fully used in the construction. Replacing it by (20 ) we get
α
n
∀K∩Fnα (x)
x
}|
{
z
z}|{ z}|{ z}|{
( N ω × N × D ) → ((N → B) → B)) → (((D → N ) → B) → B) .
|
{z
}
|
{z
}
condition 20
conclusion
Additionally the pair α, n is really coding a finite sequence, and, as we have seen, exhaustible
sets of natural numbers are uniformly equivalent to finite enumerations of natural numbers.
Hence the above can be written as
α,n
x
z}|{ z}|{
(( N ∗ × D ) →
|
{z
∀K∩Fnα (x)
condition 20
z}|{
N ∗ ) → (((D → N ) → B) → B) .
}
|
{z
}
conclusion
Therefore the above characterization reduces the type level of ∀K by two.
The last step of the proof of this theorem mimics topological proofs of Arzela–Ascoli
type theorems (which we haven’t included): to show that K ⊆ Y X is compact
Qunder assumptions such as those of Gale’s theorem (Section 7.1), one first concludes that x∈X K(x)
is compact by the Tychonoff theorem, then shows that the relative topology of K is the
topology of pointwise convergence, and that it is pointwise closed, and hence concludes that
it is homeomorphically embedded into the product as a closed subset, and therefore that
it must be compact. In the proof below, we have replaced the Tychonoff theorem by its
countable computational version given by Theorem 4.6, using a dense sequence of the exponent. The first steps of the proof are needed in order to make this replacement possible,
and they are modifications of the constructions developed in Section 6.
Proof. (⇒) (1): Theorem 6.3.
(2): Define ∀K∩F (p) = ∀f ∈ K.ψF (f ) =⇒ p(f (x)).
(⇐): By Theorem 7.4, the set K is topologically exhaustible, and hence is Kleene–
Kreisel compact by Lemma 5.5. This compactness conclusion is our only use of Theorem 7.4
in this proof. We apply this to establish the correctness of the algorithms defined below.
Define P : (D → N ) → N ω by P (f )(i) = f (di ), as in the proof of Theorem 6.3, where
d ∈ Dω is a computable dense sequence, and let L = P (K). Because Fnα is decidable on K,
the set Knα = K ∩ Fnα is exhaustible by Proposition 4.2, and Knα (x) is exhaustible uniformly
EXHAUSTIBLE SETS
29
in α, n and x ∈ K by Proposition 4.3 applied to evaluation at x. Now modify the definition
of r : N ω → N ω given in Theorem 6.3 as follows:
(
r(α)
αn
if ∃y ∈ Kn (dn ).αn = y,
r(α)(n) =
r(α)
µy.∃y 0 ∈ Kn (dn ).y = y 0 otherwise.
Then r is computable, and satisfies
(
αn
if ∃f ∈ K.f (dn ) = αn ∧ ∀i < n.f (di ) = r(α)(i),
r(α)(n) =
µy.∃f ∈ K.f (dn ) = y ∧ ∀i < n.f (di ) = r(α)(i)
otherwise.
Hence it also satisfies
(
αn
if ∃β ∈ L.βn = αn ∧ βi =n r(α),
r(α)(n) =
µy.∃β ∈ L.βn = y ∧ β =n r(α) otherwise.
This shows that r = rL for rL as defined in Theorem 6.3. But notice that, although the
second and third equations hold, the algorithm is not the same as in Theorem 6.3. In fact,
the second and third equations don’t establish computability of r, because exhaustibility of
K and L are not known at this stage of the proof. In any case, the last equation shows that
r exhibits L as a retract up to total equivalence, using the fact that L, being the continuous
P -image of K, is topologically exhaustible and hence is Kleene–Kreisel compact, as in
Theorem 6.3
Similarly, modify the definition of E : N ω → (D → N ) in Theorem 6.3 as follows:
E(α)(x) = µy.∃y 0 ∈ Knr(α) (x).y = y 0 ,
r(α)
where n is the least number such that ∀y, y 0 ∈ Kn (x).y = y 0 . Because this condition is
r(α)
equivalent to ∀f, f 0 ∈ Kn .f (x) = f 0 (x), such a number exists by Lemma 6.4 and the
r(α)
compactness of the shadow of K. By uniform exhaustibility of the set Kn (x), this can be
found uniformly in α and x, and hence E is computable. Moreover, although the definition
of E is not the same, as before, we again have E = EL for EL defined as in Theorem 6.3.
Finally, because K = K ∩ F for FQ= (D → N ), the set K(x) is exhaustible uniformly in
x ∈ D total, and hence the set M = i K(di ) ⊆ N ω is searchable uniformly in x 7→ ∀K(x) .
In fact, each K(di ) is searchable uniformly in i, by Theorem 6.3, and hence M is searchable
by Theorem 4.6. Now L ⊆ M and hence the entire r-image of M is L, and hence L is
searchable by Proposition 4.16. In turn K is the entire E-image of L and hence is also
searchable. Therefore it is exhaustible.
Notice that the proof actually concludes that K is searchable, and hence we could
have formulated the theorem as: An entire set K ⊆ (D → N ) is searchable iff K is cosemi-decidable and the set (K ∩ F )(x) is exhaustible for any F decidable on K, and any
x ∈ D total, uniformly in F and x. But this strengthening of the theorem follows from the
given formulation and the results of Section 6. However, we could have included the above
theorem, with the stronger formulation, before Section 6 and then derived the results of
that section as a corollaries. But we feel that the developments of both sections become
more mathematically transparent with the current organization of the technical material.
30
ESCARDÓ
8. Technical remarks, further work, applications and directions
We now discuss some technical aspects of the above development, announce some results
that we intend to report elsewhere, and discuss potential applications and directions for
future work in this field.
8.1. Analysis of the selection functional given by the product functional. Using
course of values induction, one easily sees that a functional
Π : ((D → B) → D)ω → ((Dω → B) → Dω )
satisfies the equation of Definition 4.5 if and only if it satisfies the equation
Π(ε)(p) = x0 ∗ Π(ε0 )(px0 ) where x0 = ε0 (λx.px (Π(ε0 )(px )))
and where px (α) = p(x ∗ α) and ε0i = εi+1 . Now define a selection function
ε2 : (B → B) → B
for 2 ⊆ B by
ε2 (p) = p(1) = if p(1) then 1 else 0
and a selection function
δ : (B ω → B) → Bω
for the Cantor space 2ω ⊆ Bω by
δ = Π(λi.ε2 ).
Then δ satisfies the equation
δ(p) = x0 ∗ δ(px0 ) where x0 = p1 (δ(p1 )).
An interesting aspect of this selection function for the Cantor space is that it doesn’t perform
case analysis on the value of p, and so, in some sense, it doesn’t work by trial and error.
In order to understand this, first notice that the above recursive definition of δ makes
sense if the domain of booleans is replaced by any domain T with an element 1 ∈ T :
δ : (T ω → T ) → T ω .
We consider the case in which T = Tω is the domain of possibly non-well-founded ωbranching trees with leaves labelled by 1. We define this as the canonical solution of the
domain equation
T ∼
= {1} + T ω ,
where the sum is lifted. Thus, a tree is either ⊥, or else a leaf 1, or else an unlabelled root
followed by a forest of countably many trees. Denote the canonical isomorphism by
[1,P ]
{1} + T ω −→ T.
Then P : T ω → T and the forest δ(P ) ∈ T ω gives a general formula for solving p(α) = 1
with α ranging over 2ω . In fact, for any given p ∈ (B ω → B), define an evaluation function
eval = evalp : T → B by
eval(1) = 1,
eval(P (α)) = p(λi. eval(αi )).
EXHAUSTIBLE SETS
31
Equivalently, eval is the unique homomorphism from the initial algebra [1, P ] : {1}+T ω → T
to the algebra [1, p] : {1} + B ω → B. Hence the solution α of the equation p(α) = 1 is given
by evaluating the general solution δ(P ) at p:
αi = evalp (δ(P )(i)).
We illustrate this with finite forests. Any p ∈ (B ω → B) defined on 2ω is uniformly
continuous, and hence of the form p(α) = q(α0 , . . . , αn−1 ) for some n and for q : B n → B
defined by this equation. Now consider the domain T = Tn of n-branching trees,
T ∼
= {1} + T n ,
and denote the canonical isomorphism by
[1,Q]
{1} + T n −→ T.
To make sense of the above definition of δ for this choice of T , define x∗(α0 , . . . , αn−2 , αn−1 ) =
(x, α0 , . . . , αn−2 ) ∈ T n for x ∈ T and α ∈ T n . We tabulate some forests, which grow doubly exponentially, but only exponentially if auxiliary variables are used to denote common
subtrees (corresponding to the variable x0 in the recursive definition of δ):
n
1
2
3
δ(Q)
Q(1)
(Q(1, Q(1, 1)), Q(Q(1, Q(1, 1)), 1))
(x0 , x1 , x2 )
where
x0 = Q(1, y, Q(1, y, 1)) with y = Q(1, 1, Q(1, 1, 1)),
x1 = Q(x0 , 1, Q(x0 , 1, 1)),
x2 = Q(x0 , x1 , 1).
In order to find (x0 , x1 , x2 ) ∈ B 3 such that q(x0 , x1 , x2 ) = 1 holds, we substitute q for Q
in the above equations, compute (x0 , x1 , x2 ), and check whether q(x0 , x1 , x2 ) = 1 holds. If
it does, then we have found a solution (in fact the largest in the lexicographic order), and
otherwise we conclude that there is no solution. Thus, the forest δ(Q) gives a closed formula
for solving the equation q(α) = 1, and telling whether there is a solution, composed only
from q and the constant 1. To solve q(α) = 0, just replace 1 by 0 in the formula.
8.2. Solution of equations with exhaustible domain. By definition, a set K ⊆ D is
searchable iff for every predicate p ∈ (D → B) defined on K one can find x0 ∈ K, uniformly
in p, such that if the equation p(x) = 1 has a solution x ∈ K, then x = x0 is a solution.
We first observe that this is equivalent to requiring that for every function f ∈ (D → N )
defined on K and any total y ∈ N one can find x0 ∈ K, uniformly in f and y, such that if
the equation f (x) = y has a solution x ∈ K, then x = x0 is a solution. For one direction,
consider the predicate p(x) = (f (x) == y), and, for the other, consider the natural inclusion
of B into N (which is the identity under our notation). Clearly, this generalizes from N to
any domain E = Dσ with σ discrete in the sense of Definition 4.12. But notice that in this
case the equation has to be written in the form f (x) ∼ y.
32
ESCARDÓ
It is natural to ask whether this generalizes to functions f ∈ (D → E) with E arbitrary.
But it is known that, in general, if an equation has more than one solution, it is typically
not possible to algorithmically find some solution [6]. We announce the following result:
Let D = Dσ→ι and E = Dτ →ι for types σ and τ , let K ⊆ D be an exhaustible
entire set, F ∈ (D → E) be total and g0 ∈ E be total.
(1) If the equation F (f ) ∼ g0 has a solution f ∈ K, unique up to total
equivalence, then some f0 ∼ f is computable, uniformly in F , g0 and
any universal quantification functional for K.
(2) It is semi-decidable whether F (f ) ∼ g0 doesn’t have a solution f ∈ K,
with the same uniformity condition.
In order to establish this, we prove the following generalization ofTLemma 6.4: Let Kn ⊆
Dσ→ι be a sequence of entire sets such that Kn ⊇ Kn+1 and that n Kn is the equivalence
class of some total f . Then one can find a computable total function f0 ∼ f , uniformly in
any sequence of universal quantification functionals for Kn .
This is not very useful in computable analysis via representations, because typically
uniqueness, when it holds, is only up to equivalence of representations rather than total
equivalence. But we do have a corresponding result for equations involving real numbers.
In light of the following, it is natural to ask whether there is a further corresponding result
for real valued functions of real variables.
8.3. An exhaustible set of analytic functions. An application of the exhaustibility of
the Cantor space to the computation of definite integrals and function maxima has been
given by Simpson [43]. A generalization of this is developed by Scriven [42]. We consider
computation with real numbers via admissible Baire-space representations [46] and domain
representations
[15]. For any x ∈ I = [−1/2, 1/2] and any sequence a ∈ [−b, b]ω , the Taylor
P
n
series n an x converges to a number in the interval [−2b, 2b]. We announce the following
example of a searchable, and hence exhaustible, set:
For any real number b > 0, the set A = Ab of analytic functions f : I → R
X
f (x) =
an xn
with a ∈ [−b, b]ω
n
has a searchable set of representatives, uniformly in b.
In our proof of this, we argue that any f ∈ A can be computed uniformly in its Taylor
coefficients and use the fact that [−b, b]ω has a searchable set of representatives. This can
be used to deduce that:
(1) The Taylor coefficients of any f ∈ A can be computed uniformly in f .
(2) The distance function dA : RI → R defined by
dA (g) = min{d(f, g) | f ∈ A}
is computable (cf. Bishop’s notion of locatedness [13, 14]).
(3) For any f ∈ RI , it is semi-decidable, uniformly in f , whether f 6∈ A.
EXHAUSTIBLE SETS
33
8.4. Peano’s theorem. This celebrated theorem asserts that certain differential equations
have solutions, but without indicating what the solutions might look like. Its proofs are
typically based on the Arzela–Ascoli theorem, and proceed by applying Euler’s algorithm to
produce a sequence of approximate solutions. In general, however, this sequence is not convergent, but, by an application of compactness, there is a convergent subsequence, although
no specific example is exhibited by this argument, which is then easily seen to produce a
solution of the equation. It is therefore natural to ask whether our tools could be applied to
compute unique solutions of such differential equations under suitable assumptions. Here
the goal is not to obtain a usable algorithm, but rather to understand the classical proof
from a computational perspective in connection with the notion of exhaustibility and its
interaction with the notion of compactness and with the Arzela–Ascoli theorem.
8.5. Uncountable products of searchable sets. It is natural to ask whether the countable product theorem 4.6 can be generalized to uncountable index sets. This question is
pertinent in view of well known constructive versions of the Tychonoff theorem in locale
theory [27] and formal topology [16], which don’t restrict the cardinality of the index set.
However, this seems unlikely in the realm of Kleene–Kreisel higher type computability theory. Consider the case in which the index set is the Cantor space. By the classical Tychonoff
theorem, the product of 2ω -many copies of 2 is compact. This product could be written
ω
as 22 . But this notation in higher-ype computation is interpreted as a function space,
ω
and in the category of Kleene–Kreisel spaces one has 22 ∼
= N, because the base is discrete
and the exponent is compact (cf. Theorem 4.13). This phenomenon in fact also takes place
in the categories of locales [26] and topological spaces [18]. In the Tychonoff theorem for
locales or spaces, the indices form a set or equivalently a discrete space. But a discrete
Kleene–Kreisel space is countable (and more generally a discrete QCB space is countable).
8.6. Totality of the product functional and bar recursion.
In Theorem 4.6 we conQ
structed a computable functional Π such that
Π(ε)(p)
∈
K
whenever
εi is a selection
i i
Q
functional for a set Ki and p is defined on i Ki . It is natural to ask whether the functional Π is actually total. Paulo Oliva has shown that this is indeed the case (personal
communication). Moreover, he has observed that if the type of booleans is replaced by the
type of natural numbers, our recursive definition of Π still makes sense and that it also
gives rise to a total functional, which he calls CBR (course-of-values bar recursion). He
additionally proved that CBR is primitively recursively inter-definable with the modified
bar recursion functional MBR defined in [11]. We are currently investigating together the
ramifications of these observations.
8.7. Alternative notions of exhaustibility. If one is interested only in total functionals and sets of total elements, it is natural formulate the following alternative notion of
exhaustibility: A set K ⊆ D is entirely exhaustible if it is entire and there is a total computable functional ∀K : (D → B) → B) such that for every total p ∈ (D → B) one has
∀K (p) = 1 iff p(x) = 1 for all x ∈ K. Because, as we have seen, non-empty, exhaustible
entire sets are computable retracts, it follows that any exhaustible entire set is entirely
exhaustible. The converse fails (but see the next paragraph), because e.g. any dense subset
of the Cantor space is entirely exhaustible using Berger’s algorithm and the fact that any
34
ESCARDÓ
total predicate is uniquely determined, up to total equivalence, by its behaviour on a dense
of set of total elements.
Moreover, when one is only interested in total functions and total elements, it is perhaps
more natural to work with Kleene–Kreisel spaces directly, without the detour via domains,
e.g. defined as k-spaces. QCB spaces are a natural and general setting for such considerations [3, 4]. One might say that a subset K of a space X is totally exhaustible if there is a
computable functional ∀K : (X → 2) → 2 such that for every p ∈ (X → 2), we have that
∀K (p) = 1 iff p(x) = 1 for all x ∈ K. When e.g. X = NN , total exhaustibility of K ⊆ X
doesn’t entail compactness of K, again considering the example of a dense subset of the
Cantor space. But Matthias Schröeder (personal communication in 2006) proved that if X
is a QCB space which is the sequential coreflection of a zero-dimensional Hausdorff space,
then any totally exhaustible closed set K ⊆ X is compact. This includes the case in which
X is a Kleene–Kreisel space. Using this and the above observations, one can show that, as
far as higher-type computation with total continuous functionals is concerned, the notions
of exhaustibility and total exhaustibility agree for closed sets.
8.8. A unified type system for total and partial computation. As we have already
discussed, Kleene–Kreisel spaces and Ershov–Scott domains live together in the cartesian
closed category of compactly generated spaces, and in fact in the subcategory of QCB spaces.
Additionally, the inclusions of k-spaces and of Ershov–Scott domains into these categories
preserve the cartesian-closed structure [21, 3]. Hence total and partial higher-type functionals coexist in the same cartesian closed category. One can envisage a higher-type system
that simultaneously incorporates, but explicitly distinguishes, total and partial objects,
and corresponding PCF-style formal systems. Among the formation rules one can have two
types for the natural numbers, with and without ⊥, and it would make sense to stipulate
that σ → τ is a partial type whenever σ is any type and τ is a partial type, and that σ → τ
is a total type when both σ and τ are total types. In its simplest form, such a language
could include Gödel’s system T for total types and PCF for partial types. Such a formalism
would have simplified, and made more transparent, much of the development concerning
exhaustible sets of total elements, where we could have benefited from functionals that take
total inputs and produce potentially partial outputs. In particular, all the technical considerations of total equivalence and shadows could have been avoided in this way, making the
development more transparent. Such functionals are actually total, but their construction
uses modes of definition that belong to the realm of partial computation. The system-T
fragment could be further extended with total computable functionals such as bar recursion
and some of those developed here, once one has shown they are indeed total.
8.9. Time complexity of exhaustive search. In the paper [19], we report some surprisingly fast experimental results, which serve to counteract an impression that might be
gained from the technical development that the algorithms presented would be essentially
intractable and of purely theoretical interest. Moreover, that paper formulates run-time
conjectures that provide examples of questions that one would like to be able to treat rigorously and that are potentially useful as target problems for work in higher-type complexity
theory. The conjectures express the run time in terms of the modulus of uniform continuity
of the input predicate on the exhaustible set, and hence topology seems to play a role in
higher-type complexity too.
EXHAUSTIBLE SETS
35
It might be possible to apply our search algorithms to practical problems, e.g. in real
analysis and in program verification. But it is more likely that, in order to obtain feasible
algorithms, such applications will need to rely on the development of particular algorithms
for particular kinds of infinite search tasks, perhaps inspired or guided by the general
algorithms we have developed, but in any case needing new insights and techniques. In fact,
this is already the case for finite search problems, as is well known. But the fast examples
reported in [19] do highlight that the task of obtaining particular search algorithms that are
efficient for particular kinds of infinite search problems of interest is a direction of research
that deserves attention and is likely to be fruitful, and that a study of feasible infinite search
problems cries to be carried out.
8.10. A fast product functional. We have just discussed that one should look for efficient
search algorithms for specialized problems. But it is still interesting to ask how fast a general
infinite search algorithm can be. We don’t know the answer, but we report an algorithm
that outperforms all the algorithms applied for the experimental results of [19], and whose
theoretical run-time behaviour remains to be investigated.
We regard an infinite sequence t as an infinite binarily branching tree with the elements
of the sequence organized in a breadth-first manner: the root is t0 , and the left and right
branches of the node tn are t2n+1 and t2n+2 . With this in mind, define functions
root : E ω → E, left, right : E ω → E ω , branch : E × E ω × E ω → E ω
by
root(t) = t0 ,
left(t) = λi.t2i+1 ,
right(t) = λi.t2i+2 ,


if i = 0,
x
branch(x, l, r) = λi. l(i−1)/2 if i is odd,


r(i−2)/2 otherwise.
ω
Then, for x ∈ E and l, r ∈ E ,
root(branch(x, l, r)) = x,
left(branch(x, l, r)) = l,
right(branch(x, l, r)) = r,
branch(root(t), left(t), right(t)) = t.
Our experimentally faster product algorithm is then recursively defined by
Π(ε)(p) = branch(x0 , l0 , r0 )
where
εroot = root(ε),
εleft = Π(left(ε)),
εright = Π(right(ε)),
∃left (p) = p(εleft (p)),
∃right (p) = p(εright (p)),
x0 = εroot (λx.∃left l. ∃right r. p(branch(x, l, r))),
l0 = εleft (λl.∃right r. p(branch(x0 , l, r))),
r0 = εright (λr.p(branch(x0 , l0 , r)).
The idea is that treating sequences as trees reduces some linear factors to logarithmic factors
(very much like in the well-known heap-sort algorithm).
36
ESCARDÓ
8.11. Operational perspective. An advantage of the proof of Theorem 4.6 sketched
in [19] is that it can be directly interpreted in the operational setting [18, 20]. The proofs of
the other results of Section 4 are also easily seen to work in the above operational setting.
But a development of operational counter-parts for those of later sections is left as an open
problem. This requires an operational reworking of the topological Section 5, which seems
challenging.
9. Concluding remark on the role of topology
The algorithms developed in this work have purely computational specifications, which
allow them to be applied without knowledge of specialized mathematical techniques in the
theory of computation. However, the correctness proofs of some of the algorithms crucially
rely on topological techniques. In this sense, this work is a genuine application of topology
to computation: theorems formulated in the language of computation, proofs developed in
the language of topology.
But there is another sense in which topology proves to play a crucial role. Compact sets
in topology are advertised as sets that behave, in many important respects, as if they were
finite. Then exhaustively searchable sets ought to be compact. And compact sets are known
to be closed under continuous images and under finite and infinite products. Moreover, for
countably based Hausdorff spaces, they are the continuous images of the Cantor space.
Hence searchable sets ought to have corresponding closure properties and characterization,
which is what this work establishes, among other things, motivated by these considerations.
Thus, in a more abstract level, topology is applied as a paradigm for discovering unforeseen
notions, algorithms and theorems in computability theory.
References
[1] S. Abramsky and A. Jung. Domain theory. In S. Abramsky, D.M. Gabbay, and T.S.E. Maibaum, editors,
Handbook of Logic in Computer Science, volume 3 of Oxford science publications, pages 1–168. 1994.
[2] H.P. Barendregt. The Lambda-Calculus: its Syntax and Semantics. North-Holland, 1984.
[3] I. Battenfeld, M. Schröder, and A. Simpson. A convenient category of domains. In Computation, meaning, and logic: articles dedicated to Gordon Plotkin, volume 172 of Electron. Notes Theor. Comput. Sci.,
pages 69–99. Elsevier, Amsterdam, 2007.
[4] A. Bauer. A relationship between equilogical spaces and type two effectivity. MLQ Math. Log. Q.,
48(suppl. 1):1–15, 2002. Dagstuhl Seminar on Computability and Complexity in Analysis, 2001.
[5] A. Bauer, M.H. Escardó, and A.K. Simpson. Comparing functional paradigms for exact real-number
computation. volume 2380 of Lect. Not. Comp. Sci., pages 488–500, 2002.
[6] M.J. Beeson. Foundations of Constructive Mathematics. Springer, 1985.
[7] U. Berger. Totale Objekte und Mengen in der Bereichstheorie. PhD thesis, Mathematisches Institut der
Universität München, 1990.
[8] U. Berger. Total sets and objects in domain theory. Ann. Pure Appl. Logic, 60(2):91–117, 1993.
[9] U. Berger. Continuous functionals of dependent and transfinite types. In Models and computability
(Leeds, 1997), volume 259 of London Math. Soc. Lecture Note Ser., pages 1–22. Cambridge Univ. Press,
Cambridge, 1999.
[10] U Berger. Density theorems for the domains-with-totality semantics of dependent types. Appl. Categ.
Structures, 7(1-2):3–30, 1999. Applications of ordered sets in computer science (Braunschweig, 1996).
[11] U. Berger and P. Oliva. Modified bar recursion. Math. Structures Comput. Sci., 16(2):163–183, 2006.
[12] R. Bird and P. Wadler. Introduction to Functional Programming. Prentice-Hall, New York, 1988.
[13] E. Bishop. Foundations of constructive analysis. McGraw-Hill Book Co., New York, 1967.
[14] E. Bishop and D. Bridges. Constructive Analysis. Springer, Berlin, 1985.
[15] J. Blanck. Domain representations of topological spaces. Theoret. Comput. Sci., 247(1-2):229–255, 2000.
EXHAUSTIBLE SETS
37
[16] T. Coquand. An intuitionistic proof of Tychonoff’s theorem. J. Symbolic Logic, 57(1):28–32, 1992.
[17] H. Egli and R.L. Constable. Computability concepts for programming languages. Theoret. Comput.
Sci., 2:133–145, 1976.
[18] M.H. Escardó. Synthetic topology of data types and classical spaces. Electron. Notes Theor. Comput.
Sci., 87:21–156, 2004.
[19] M.H. Escardó. Infinite sets that admit fast exhaustive search. In Proceedings of the 22nd Annual IEEE
Symposium on Logic In Computer Science, pages 443–452. IEEE Computer Society, 2007.
[20] M.H. Escardó and W.K. Ho. Operational domain theory and topology of a sequential programming
language. In Proceedings of the 20th Annual IEEE Symposium on Logic In Computer Science, pages
427–436, 2005.
[21] M.H. Escardó, J. Lawson, and A. Simpson. Comparing Cartesian closed categories of (core) compactly
generated spaces. Topology Appl., 143(1-3):105–145, 2004.
[22] D. Gale. Compact sets of functions and function rings. Proc. Amer. Math. Soc., 1:303–308, 1950.
[23] R. O. Gandy and J. M. E. Hyland. Computable and recursively countable functions of higher type.
In Logic Colloquium 76 (Oxford, 1976), pages 407–438. Studies in Logic and Found. Math., Vol. 87.
North-Holland, Amsterdam, 1977.
[24] G. Gierz, K.H. Hofmann, K. Keimel, J.D. Lawson, M. Mislove, and D.S. Scott. Continuous Lattices and
Domains. Cambridge University Press, 2003.
[25] G. Hutton. Programming in Haskell. Cambridge University Press, 2007.
[26] M. Hyland. Function spaces in the category of locales. In Continuous lattices, volume 871 of Lect. Notes
Math., pages 264–281, 1981.
[27] P. T. Johnstone. Tychonoff’s theorem without the axiom of choice. Fund. Math., 113(1):21–35, 1981.
[28] J.L. Kelley. General Topology. D. van Nostrand, New York, 1955.
[29] J.R. Longley. Notions of computability at higher types. I. In Logic Colloquium 2000, volume 19 of Lect.
Notes Log., pages 32–142. Assoc. Symbol. Logic, Urbana, IL, 2005.
[30] J.R. Longley. On the ubiquity of certain type structures. Mathematical Structures in Computer Science,
17:841–953, 2007.
[31] D. Normann. Recursion on the countable functionals, volume 811 of Lec. Not. Math. Springer, 1980.
[32] D. Normann. Computability over the partial continuous functionals. J. Symbolic Logic, 65(3):1133–1142,
2000.
[33] D. Normann. Comparing hierarchies of total functionals. Log. Methods Comput. Sci., 1(2):2:4, 28, 2005.
[34] D. Normann. Computing with functionals—computability theory or computer science? Bull. Symbolic
Logic, 12(1):43–59, 2006.
[35] G.D. Plotkin. LCF considered as a programming language. Theoret. Comput. Sci., 5(1):223–255, 1977.
[36] G.D. Plotkin. Tω as a universal domain. J. Comput. System Sci., 17:209–236, 1978.
[37] G.D. Plotkin. Pisa notes on domains. Department of Computer Science, University of Edinburgh. Available at the author’s web page, 1983.
[38] G.D. Plotkin. Full abstraction, totality and PCF. Math. Structures Comput. Sci., 9(1):1–20, 1999.
N
[39] M. Schöder. The sequential topology on NN is not regular. Preprint. Institut für Theoretische Informatik
und Mathematik, Fakultät für Informatik, Universität der Bundeswehr München, May 2008.
[40] D.S. Scott. Data types as lattices. SIAM J. Comput., 5:522–587, 1976.
[41] D.S. Scott. A type-theoretical alternative to CUCH, ISWIM and OWHY. Theoret. Comput. Sci.,
121:411–440, 1993. Reprint of a 1969 manuscript.
[42] A. Scriven. A functional algorithm for exact real integration with invariant measures. In Mathematical
Foundations of Programming Semantics, page To appear, 2008. Electr. Notes. Theret. Comp. Sci.
[43] A. Simpson. Lazy functional algorithms for exact real functionals. Lec. Not. Comput. Sci., 1450:323–342,
1998.
[44] M.B. Smyth. Effectively given domains. Theoret. Comput. Sci., 5(1):256–274, 1977.
[45] M.B. Smyth. Topology. In S. Abramsky, D.M. Gabbay, and T.S.E. Maibaum, editors, Handbook of Logic
in Computer Science, volume 1 of Oxford science publications, pages 641–761. 1992.
[46] K. Weihrauch. Computable analysis. Springer, 2000.